1 /* 2 * Copyright 2007 Sun Microsystems, Inc. All rights reserved. 3 * Use is subject to license terms. 4 * 5 * Copyright (c) 1983, 1988, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgment: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * $FreeBSD: src/sbin/routed/table.c,v 1.15 2000/08/11 08:24:38 sheldonh Exp $ 37 */ 38 39 #pragma ident "%Z%%M% %I% %E% SMI" 40 41 #include "defs.h" 42 #include <fcntl.h> 43 #include <stropts.h> 44 #include <sys/tihdr.h> 45 #include <inet/mib2.h> 46 #include <inet/ip.h> 47 48 /* This structure is used to store a disassembled routing socket message. */ 49 struct rt_addrinfo { 50 int rti_addrs; 51 struct sockaddr_storage *rti_info[RTAX_MAX]; 52 }; 53 54 static struct rt_spare *rts_better(struct rt_entry *); 55 static struct rt_spare rts_empty = EMPTY_RT_SPARE; 56 static void set_need_flash(void); 57 static void rtbad(struct rt_entry *, struct interface *); 58 static int rt_xaddrs(struct rt_addrinfo *, struct sockaddr_storage *, 59 char *, int); 60 static struct interface *gwkludge_iflookup(in_addr_t, in_addr_t, in_addr_t); 61 62 struct radix_node_head *rhead; /* root of the radix tree */ 63 64 /* Flash update needed. _B_TRUE to suppress the 1st. */ 65 boolean_t need_flash = _B_TRUE; 66 67 struct timeval age_timer; /* next check of old routes */ 68 struct timeval need_kern = { /* need to update kernel table */ 69 EPOCH+MIN_WAITTIME-1, 0 70 }; 71 72 static uint32_t total_routes; 73 74 #define ROUNDUP_LONG(a) \ 75 ((a) > 0 ? (1 + (((a) - 1) | (sizeof (long) - 1))) : sizeof (long)) 76 77 /* 78 * It is desirable to "aggregate" routes, to combine differing routes of 79 * the same metric and next hop into a common route with a smaller netmask 80 * or to suppress redundant routes, routes that add no information to 81 * routes with smaller netmasks. 82 * 83 * A route is redundant if and only if any and all routes with smaller 84 * but matching netmasks and nets are the same. Since routes are 85 * kept sorted in the radix tree, redundant routes always come second. 86 * 87 * There are two kinds of aggregations. First, two routes of the same bit 88 * mask and differing only in the least significant bit of the network 89 * number can be combined into a single route with a coarser mask. 90 * 91 * Second, a route can be suppressed in favor of another route with a more 92 * coarse mask provided no incompatible routes with intermediate masks 93 * are present. The second kind of aggregation involves suppressing routes. 94 * A route must not be suppressed if an incompatible route exists with 95 * an intermediate mask, since the suppressed route would be covered 96 * by the intermediate. 97 * 98 * This code relies on the radix tree walk encountering routes 99 * sorted first by address, with the smallest address first. 100 */ 101 102 static struct ag_info ag_slots[NUM_AG_SLOTS], *ag_avail, *ag_corsest, 103 *ag_finest; 104 105 #ifdef DEBUG_AG 106 #define CHECK_AG() do { int acnt = 0; struct ag_info *cag; \ 107 for (cag = ag_avail; cag != NULL; cag = cag->ag_fine) \ 108 acnt++; \ 109 for (cag = ag_corsest; cag != NULL; cag = cag->ag_fine) \ 110 acnt++; \ 111 if (acnt != NUM_AG_SLOTS) \ 112 abort(); \ 113 } while (_B_FALSE) 114 #else 115 #define CHECK_AG() (void)0 116 #endif 117 118 119 /* 120 * Output the contents of an aggregation table slot. 121 * This function must always be immediately followed with the deletion 122 * of the target slot. 123 */ 124 static void 125 ag_out(struct ag_info *ag, void (*out)(struct ag_info *)) 126 { 127 struct ag_info *ag_cors; 128 uint32_t bit; 129 130 131 /* Forget it if this route should not be output for split-horizon. */ 132 if (ag->ag_state & AGS_SPLIT_HZ) 133 return; 134 135 /* 136 * If we output both the even and odd twins, then the immediate parent, 137 * if it is present, is redundant, unless the parent manages to 138 * aggregate into something coarser. 139 * On successive calls, this code detects the even and odd twins, 140 * and marks the parent. 141 * 142 * Note that the order in which the radix tree code emits routes 143 * ensures that the twins are seen before the parent is emitted. 144 */ 145 ag_cors = ag->ag_cors; 146 if (ag_cors != NULL && 147 ag_cors->ag_mask == (ag->ag_mask << 1) && 148 ag_cors->ag_dst_h == (ag->ag_dst_h & ag_cors->ag_mask)) { 149 ag_cors->ag_state |= ((ag_cors->ag_dst_h == ag->ag_dst_h) ? 150 AGS_REDUN0 : AGS_REDUN1); 151 } 152 153 /* 154 * Skip it if this route is itself redundant. 155 * 156 * It is ok to change the contents of the slot here, since it is 157 * always deleted next. 158 */ 159 if (ag->ag_state & AGS_REDUN0) { 160 if (ag->ag_state & AGS_REDUN1) 161 return; /* quit if fully redundant */ 162 /* make it finer if it is half-redundant */ 163 bit = (-ag->ag_mask) >> 1; 164 ag->ag_dst_h |= bit; 165 ag->ag_mask |= bit; 166 167 } else if (ag->ag_state & AGS_REDUN1) { 168 /* make it finer if it is half-redundant */ 169 bit = (-ag->ag_mask) >> 1; 170 ag->ag_mask |= bit; 171 } 172 out(ag); 173 } 174 175 176 static void 177 ag_del(struct ag_info *ag) 178 { 179 CHECK_AG(); 180 181 if (ag->ag_cors == NULL) 182 ag_corsest = ag->ag_fine; 183 else 184 ag->ag_cors->ag_fine = ag->ag_fine; 185 186 if (ag->ag_fine == NULL) 187 ag_finest = ag->ag_cors; 188 else 189 ag->ag_fine->ag_cors = ag->ag_cors; 190 191 ag->ag_fine = ag_avail; 192 ag_avail = ag; 193 194 CHECK_AG(); 195 } 196 197 198 /* Look for a route that can suppress the given route. */ 199 static struct ag_info * 200 ag_find_suppressor(struct ag_info *ag) 201 { 202 struct ag_info *ag_cors; 203 in_addr_t dst_h = ag->ag_dst_h; 204 205 for (ag_cors = ag->ag_cors; ag_cors != NULL; 206 ag_cors = ag_cors->ag_cors) { 207 208 if ((dst_h & ag_cors->ag_mask) == ag_cors->ag_dst_h) { 209 /* 210 * We found a route with a coarser mask that covers 211 * the given target. It can suppress the target 212 * only if it has a good enough metric and it 213 * either has the same (gateway, ifp), or if its state 214 * includes AGS_CORS_GATE or the target's state 215 * includes AGS_FINE_GATE. 216 */ 217 if (ag_cors->ag_pref <= ag->ag_pref && 218 (((ag->ag_nhop == ag_cors->ag_nhop) && 219 (ag->ag_ifp == ag_cors->ag_ifp)) || 220 ag_cors->ag_state & AGS_CORS_GATE || 221 ag->ag_state & AGS_FINE_GATE)) { 222 return (ag_cors); 223 } 224 } 225 } 226 227 return (NULL); 228 } 229 230 231 /* 232 * Flush routes waiting for aggregation. 233 * This must not suppress a route unless it is known that among all routes 234 * with coarser masks that match it, the one with the longest mask is 235 * appropriate. This is ensured by scanning the routes in lexical order, 236 * and with the most restrictive mask first among routes to the same 237 * destination. 238 */ 239 void 240 ag_flush(in_addr_t lim_dst_h, /* flush routes to here */ 241 in_addr_t lim_mask, /* matching this mask */ 242 void (*out)(struct ag_info *)) 243 { 244 struct ag_info *ag, *ag_cors, *ag_supr; 245 in_addr_t dst_h; 246 247 248 for (ag = ag_finest; ag != NULL && ag->ag_mask >= lim_mask; 249 ag = ag_cors) { 250 /* Get the next route now, before we delete ag. */ 251 ag_cors = ag->ag_cors; 252 253 /* Work on only the specified routes. */ 254 dst_h = ag->ag_dst_h; 255 if ((dst_h & lim_mask) != lim_dst_h) 256 continue; 257 258 /* 259 * Don't try to suppress the route if its state doesn't 260 * include AGS_SUPPRESS. 261 */ 262 if (!(ag->ag_state & AGS_SUPPRESS)) { 263 ag_out(ag, out); 264 ag_del(ag); 265 continue; 266 } 267 268 ag_supr = ag_find_suppressor(ag); 269 if (ag_supr == NULL) { 270 /* 271 * We didn't find a route which suppresses the 272 * target, so the target can go out. 273 */ 274 ag_out(ag, out); 275 } else { 276 /* 277 * We found a route which suppresses the target, so 278 * don't output the target. 279 */ 280 if (TRACEACTIONS) { 281 trace_misc("aggregated away %s", 282 rtname(htonl(ag->ag_dst_h), ag->ag_mask, 283 ag->ag_nhop)); 284 trace_misc("on coarser route %s", 285 rtname(htonl(ag_supr->ag_dst_h), 286 ag_supr->ag_mask, ag_supr->ag_nhop)); 287 } 288 /* 289 * If the suppressed target was redundant, then 290 * mark the suppressor as redundant. 291 */ 292 if (AG_IS_REDUN(ag->ag_state) && 293 ag_supr->ag_mask == (ag->ag_mask<<1)) { 294 if (ag_supr->ag_dst_h == dst_h) 295 ag_supr->ag_state |= AGS_REDUN0; 296 else 297 ag_supr->ag_state |= AGS_REDUN1; 298 } 299 if (ag->ag_tag != ag_supr->ag_tag) 300 ag_supr->ag_tag = 0; 301 if (ag->ag_nhop != ag_supr->ag_nhop) 302 ag_supr->ag_nhop = 0; 303 } 304 305 /* The route has either been output or suppressed */ 306 ag_del(ag); 307 } 308 309 CHECK_AG(); 310 } 311 312 313 /* Try to aggregate a route with previous routes. */ 314 void 315 ag_check(in_addr_t dst, 316 in_addr_t mask, 317 in_addr_t gate, 318 struct interface *ifp, 319 in_addr_t nhop, 320 uint8_t metric, 321 uint8_t pref, 322 uint32_t seqno, 323 uint16_t tag, 324 uint16_t state, 325 void (*out)(struct ag_info *)) /* output using this */ 326 { 327 struct ag_info *ag, *nag, *ag_cors; 328 in_addr_t xaddr; 329 int tmp; 330 struct interface *xifp; 331 332 dst = ntohl(dst); 333 334 /* 335 * Don't bother trying to aggregate routes with non-contiguous 336 * subnet masks. 337 * 338 * (X & -X) contains a single bit if and only if X is a power of 2. 339 * (X + (X & -X)) == 0 if and only if X is a power of 2. 340 */ 341 if ((mask & -mask) + mask != 0) { 342 struct ag_info nc_ag; 343 344 nc_ag.ag_dst_h = dst; 345 nc_ag.ag_mask = mask; 346 nc_ag.ag_gate = gate; 347 nc_ag.ag_ifp = ifp; 348 nc_ag.ag_nhop = nhop; 349 nc_ag.ag_metric = metric; 350 nc_ag.ag_pref = pref; 351 nc_ag.ag_tag = tag; 352 nc_ag.ag_state = state; 353 nc_ag.ag_seqno = seqno; 354 out(&nc_ag); 355 return; 356 } 357 358 /* Search for the right slot in the aggregation table. */ 359 ag_cors = NULL; 360 ag = ag_corsest; 361 while (ag != NULL) { 362 if (ag->ag_mask >= mask) 363 break; 364 365 /* 366 * Suppress old routes (i.e. combine with compatible routes 367 * with coarser masks) as we look for the right slot in the 368 * aggregation table for the new route. 369 * A route to an address less than the current destination 370 * will not be affected by the current route or any route 371 * seen hereafter. That means it is safe to suppress it. 372 * This check keeps poor routes (e.g. with large hop counts) 373 * from preventing suppression of finer routes. 374 */ 375 if (ag_cors != NULL && ag->ag_dst_h < dst && 376 (ag->ag_state & AGS_SUPPRESS) && 377 ag_cors->ag_pref <= ag->ag_pref && 378 (ag->ag_dst_h & ag_cors->ag_mask) == ag_cors->ag_dst_h && 379 ((ag_cors->ag_nhop == ag->ag_nhop && 380 (ag_cors->ag_ifp == ag->ag_ifp))|| 381 (ag->ag_state & AGS_FINE_GATE) || 382 (ag_cors->ag_state & AGS_CORS_GATE))) { 383 /* 384 * If the suppressed target was redundant, 385 * then mark the suppressor redundant. 386 */ 387 if (AG_IS_REDUN(ag->ag_state) && 388 ag_cors->ag_mask == (ag->ag_mask << 1)) { 389 if (ag_cors->ag_dst_h == dst) 390 ag_cors->ag_state |= AGS_REDUN0; 391 else 392 ag_cors->ag_state |= AGS_REDUN1; 393 } 394 if (ag->ag_tag != ag_cors->ag_tag) 395 ag_cors->ag_tag = 0; 396 if (ag->ag_nhop != ag_cors->ag_nhop) 397 ag_cors->ag_nhop = 0; 398 ag_del(ag); 399 CHECK_AG(); 400 } else { 401 ag_cors = ag; 402 } 403 ag = ag_cors->ag_fine; 404 } 405 406 /* 407 * If we find the even/odd twin of the new route, and if the 408 * masks and so forth are equal, we can aggregate them. 409 * We can probably promote one of the pair. 410 * 411 * Since the routes are encountered in lexical order, 412 * the new route must be odd. However, the second or later 413 * times around this loop, it could be the even twin promoted 414 * from the even/odd pair of twins of the finer route. 415 */ 416 while (ag != NULL && ag->ag_mask == mask && 417 ((ag->ag_dst_h ^ dst) & (mask<<1)) == 0) { 418 419 /* 420 * Here we know the target route and the route in the current 421 * slot have the same netmasks and differ by at most the 422 * last bit. They are either for the same destination, or 423 * for an even/odd pair of destinations. 424 */ 425 if (ag->ag_dst_h == dst) { 426 if (ag->ag_nhop == nhop && ag->ag_ifp == ifp) { 427 /* 428 * We have two routes to the same destination, 429 * with the same nexthop and interface. 430 * Routes are encountered in lexical order, 431 * so a route is never promoted until the 432 * parent route is already present. So we 433 * know that the new route is a promoted (or 434 * aggregated) pair and the route already in 435 * the slot is the explicit route. 436 * 437 * Prefer the best route if their metrics 438 * differ, or the aggregated one if not, 439 * following a sort of longest-match rule. 440 */ 441 if (pref <= ag->ag_pref) { 442 ag->ag_gate = gate; 443 ag->ag_ifp = ifp; 444 ag->ag_nhop = nhop; 445 ag->ag_tag = tag; 446 ag->ag_metric = metric; 447 ag->ag_pref = pref; 448 if (seqno > ag->ag_seqno) 449 ag->ag_seqno = seqno; 450 tmp = ag->ag_state; 451 ag->ag_state = state; 452 state = tmp; 453 } 454 455 /* 456 * Some bits are set if they are set on 457 * either route, except when the route is 458 * for an interface. 459 */ 460 if (!(ag->ag_state & AGS_IF)) 461 ag->ag_state |= 462 (state & (AGS_AGGREGATE_EITHER | 463 AGS_REDUN0 | AGS_REDUN1)); 464 465 return; 466 } else { 467 /* 468 * multiple routes to same dest/mask with 469 * differing gate nexthop/or ifp. Flush 470 * both out. 471 */ 472 break; 473 } 474 } 475 476 /* 477 * If one of the routes can be promoted and the other can 478 * be suppressed, it may be possible to combine them or 479 * worthwhile to promote one. 480 * 481 * Any route that can be promoted is always 482 * marked to be eligible to be suppressed. 483 */ 484 if (!((state & AGS_AGGREGATE) && 485 (ag->ag_state & AGS_SUPPRESS)) && 486 !((ag->ag_state & AGS_AGGREGATE) && (state & AGS_SUPPRESS))) 487 break; 488 489 /* 490 * A pair of even/odd twin routes can be combined 491 * if either is redundant, or if they are via the 492 * same gateway and have the same metric. 493 */ 494 if (AG_IS_REDUN(ag->ag_state) || AG_IS_REDUN(state) || 495 (ag->ag_nhop == nhop && ag->ag_ifp == ifp && 496 ag->ag_pref == pref && 497 (state & ag->ag_state & AGS_AGGREGATE) != 0)) { 498 499 /* 500 * We have both the even and odd pairs. 501 * Since the routes are encountered in order, 502 * the route in the slot must be the even twin. 503 * 504 * Combine and promote (aggregate) the pair of routes. 505 */ 506 if (seqno < ag->ag_seqno) 507 seqno = ag->ag_seqno; 508 if (!AG_IS_REDUN(state)) 509 state &= ~AGS_REDUN1; 510 if (AG_IS_REDUN(ag->ag_state)) 511 state |= AGS_REDUN0; 512 else 513 state &= ~AGS_REDUN0; 514 state |= (ag->ag_state & AGS_AGGREGATE_EITHER); 515 if (ag->ag_tag != tag) 516 tag = 0; 517 if (ag->ag_nhop != nhop) 518 nhop = 0; 519 520 /* 521 * Get rid of the even twin that was already 522 * in the slot. 523 */ 524 ag_del(ag); 525 526 } else if (ag->ag_pref >= pref && 527 (ag->ag_state & AGS_AGGREGATE)) { 528 /* 529 * If we cannot combine the pair, maybe the route 530 * with the worse metric can be promoted. 531 * 532 * Promote the old, even twin, by giving its slot 533 * in the table to the new, odd twin. 534 */ 535 ag->ag_dst_h = dst; 536 537 xaddr = ag->ag_gate; 538 ag->ag_gate = gate; 539 gate = xaddr; 540 541 xifp = ag->ag_ifp; 542 ag->ag_ifp = ifp; 543 ifp = xifp; 544 545 xaddr = ag->ag_nhop; 546 ag->ag_nhop = nhop; 547 nhop = xaddr; 548 549 tmp = ag->ag_tag; 550 ag->ag_tag = tag; 551 tag = tmp; 552 553 /* 554 * The promoted route is even-redundant only if the 555 * even twin was fully redundant. It is not 556 * odd-redundant because the odd-twin will still be 557 * in the table. 558 */ 559 tmp = ag->ag_state; 560 if (!AG_IS_REDUN(tmp)) 561 tmp &= ~AGS_REDUN0; 562 tmp &= ~AGS_REDUN1; 563 ag->ag_state = state; 564 state = tmp; 565 566 tmp = ag->ag_metric; 567 ag->ag_metric = metric; 568 metric = tmp; 569 570 tmp = ag->ag_pref; 571 ag->ag_pref = pref; 572 pref = tmp; 573 574 /* take the newest sequence number */ 575 if (seqno <= ag->ag_seqno) 576 seqno = ag->ag_seqno; 577 else 578 ag->ag_seqno = seqno; 579 580 } else { 581 if (!(state & AGS_AGGREGATE)) 582 break; /* cannot promote either twin */ 583 584 /* 585 * Promote the new, odd twin by shaving its 586 * mask and address. 587 * The promoted route is odd-redundant only if the 588 * odd twin was fully redundant. It is not 589 * even-redundant because the even twin is still in 590 * the table. 591 */ 592 if (!AG_IS_REDUN(state)) 593 state &= ~AGS_REDUN1; 594 state &= ~AGS_REDUN0; 595 if (seqno < ag->ag_seqno) 596 seqno = ag->ag_seqno; 597 else 598 ag->ag_seqno = seqno; 599 } 600 601 mask <<= 1; 602 dst &= mask; 603 604 if (ag_cors == NULL) { 605 ag = ag_corsest; 606 break; 607 } 608 ag = ag_cors; 609 ag_cors = ag->ag_cors; 610 } 611 612 /* 613 * When we can no longer promote and combine routes, 614 * flush the old route in the target slot. Also flush 615 * any finer routes that we know will never be aggregated by 616 * the new route. 617 * 618 * In case we moved toward coarser masks, 619 * get back where we belong 620 */ 621 if (ag != NULL && ag->ag_mask < mask) { 622 ag_cors = ag; 623 ag = ag->ag_fine; 624 } 625 626 /* Empty the target slot */ 627 if (ag != NULL && ag->ag_mask == mask) { 628 ag_flush(ag->ag_dst_h, ag->ag_mask, out); 629 ag = (ag_cors == NULL) ? ag_corsest : ag_cors->ag_fine; 630 } 631 632 #ifdef DEBUG_AG 633 if (ag == NULL && ag_cors != ag_finest) 634 abort(); 635 if (ag_cors == NULL && ag != ag_corsest) 636 abort(); 637 if (ag != NULL && ag->ag_cors != ag_cors) 638 abort(); 639 if (ag_cors != NULL && ag_cors->ag_fine != ag) 640 abort(); 641 CHECK_AG(); 642 #endif 643 644 /* Save the new route on the end of the table. */ 645 nag = ag_avail; 646 ag_avail = nag->ag_fine; 647 648 nag->ag_dst_h = dst; 649 nag->ag_mask = mask; 650 nag->ag_ifp = ifp; 651 nag->ag_gate = gate; 652 nag->ag_nhop = nhop; 653 nag->ag_metric = metric; 654 nag->ag_pref = pref; 655 nag->ag_tag = tag; 656 nag->ag_state = state; 657 nag->ag_seqno = seqno; 658 659 nag->ag_fine = ag; 660 if (ag != NULL) 661 ag->ag_cors = nag; 662 else 663 ag_finest = nag; 664 nag->ag_cors = ag_cors; 665 if (ag_cors == NULL) 666 ag_corsest = nag; 667 else 668 ag_cors->ag_fine = nag; 669 CHECK_AG(); 670 } 671 672 673 static const char * 674 rtm_type_name(uchar_t type) 675 { 676 static const char *rtm_types[] = { 677 "RTM_ADD", 678 "RTM_DELETE", 679 "RTM_CHANGE", 680 "RTM_GET", 681 "RTM_LOSING", 682 "RTM_REDIRECT", 683 "RTM_MISS", 684 "RTM_LOCK", 685 "RTM_OLDADD", 686 "RTM_OLDDEL", 687 "RTM_RESOLVE", 688 "RTM_NEWADDR", 689 "RTM_DELADDR", 690 "RTM_IFINFO", 691 "RTM_NEWMADDR", 692 "RTM_DELMADDR" 693 }; 694 #define NEW_RTM_PAT "RTM type %#x" 695 static char name0[sizeof (NEW_RTM_PAT) + 2]; 696 697 if (type > sizeof (rtm_types) / sizeof (rtm_types[0]) || type == 0) { 698 (void) snprintf(name0, sizeof (name0), NEW_RTM_PAT, type); 699 return (name0); 700 } else { 701 return (rtm_types[type-1]); 702 } 703 #undef NEW_RTM_PAT 704 } 705 706 707 static void 708 dump_rt_msg(const char *act, struct rt_msghdr *rtm, int mlen) 709 { 710 const char *mtype; 711 uchar_t *cp; 712 int i, j; 713 char buffer[16*3 + 1], *ibs; 714 struct ifa_msghdr *ifam; 715 struct if_msghdr *ifm; 716 717 switch (rtm->rtm_type) { 718 case RTM_NEWADDR: 719 case RTM_DELADDR: 720 mtype = "ifam"; 721 break; 722 case RTM_IFINFO: 723 mtype = "ifm"; 724 break; 725 default: 726 mtype = "rtm"; 727 break; 728 } 729 trace_misc("%s %s %d bytes", act, mtype, mlen); 730 if (mlen > rtm->rtm_msglen) { 731 trace_misc("%s: extra %d bytes ignored", mtype, 732 mlen - rtm->rtm_msglen); 733 mlen = rtm->rtm_msglen; 734 } else if (mlen < rtm->rtm_msglen) { 735 trace_misc("%s: truncated by %d bytes", mtype, 736 rtm->rtm_msglen - mlen); 737 } 738 switch (rtm->rtm_type) { 739 case RTM_NEWADDR: 740 case RTM_DELADDR: 741 ifam = (struct ifa_msghdr *)rtm; 742 trace_misc("ifam: msglen %d version %d type %d addrs %X", 743 ifam->ifam_msglen, ifam->ifam_version, ifam->ifam_type, 744 ifam->ifam_addrs); 745 trace_misc("ifam: flags %X index %d metric %d", 746 ifam->ifam_flags, ifam->ifam_index, ifam->ifam_metric); 747 cp = (uchar_t *)(ifam + 1); 748 break; 749 case RTM_IFINFO: 750 ifm = (struct if_msghdr *)rtm; 751 trace_misc("ifm: msglen %d version %d type %d addrs %X", 752 ifm->ifm_msglen, ifm->ifm_version, ifm->ifm_type, 753 ifm->ifm_addrs); 754 ibs = if_bit_string(ifm->ifm_flags, _B_TRUE); 755 if (ibs == NULL) { 756 trace_misc("ifm: flags %#x index %d", ifm->ifm_flags, 757 ifm->ifm_index); 758 } else { 759 trace_misc("ifm: flags %s index %d", ibs, 760 ifm->ifm_index); 761 free(ibs); 762 } 763 cp = (uchar_t *)(ifm + 1); 764 break; 765 default: 766 trace_misc("rtm: msglen %d version %d type %d index %d", 767 rtm->rtm_msglen, rtm->rtm_version, rtm->rtm_type, 768 rtm->rtm_index); 769 trace_misc("rtm: flags %X addrs %X pid %d seq %d", 770 rtm->rtm_flags, rtm->rtm_addrs, rtm->rtm_pid, rtm->rtm_seq); 771 trace_misc("rtm: errno %d use %d inits %X", rtm->rtm_errno, 772 rtm->rtm_use, rtm->rtm_inits); 773 cp = (uchar_t *)(rtm + 1); 774 break; 775 } 776 i = mlen - (cp - (uint8_t *)rtm); 777 while (i > 0) { 778 buffer[0] = '\0'; 779 ibs = buffer; 780 for (j = 0; j < 16 && i > 0; j++, i--) 781 ibs += sprintf(ibs, " %02X", *cp++); 782 trace_misc("addr%s", buffer); 783 } 784 } 785 786 /* 787 * Tell the kernel to add, delete or change a route 788 * Pass k_state from khash in for diagnostic info. 789 */ 790 static void 791 rtioctl(int action, /* RTM_DELETE, etc */ 792 in_addr_t dst, 793 in_addr_t gate, 794 in_addr_t mask, 795 struct interface *ifp, 796 uint8_t metric, 797 int flags) 798 { 799 static int rt_sock_seqno = 0; 800 struct { 801 struct rt_msghdr w_rtm; 802 struct sockaddr_in w_dst; 803 struct sockaddr_in w_gate; 804 uint8_t w_space[512]; 805 } w; 806 struct sockaddr_in w_mask; 807 struct sockaddr_dl w_ifp; 808 uint8_t *cp; 809 long cc; 810 #define PAT " %-10s %s metric=%d flags=%#x" 811 #define ARGS rtm_type_name(action), rtname(dst, mask, gate), metric, flags 812 813 again: 814 (void) memset(&w, 0, sizeof (w)); 815 (void) memset(&w_mask, 0, sizeof (w_mask)); 816 (void) memset(&w_ifp, 0, sizeof (w_ifp)); 817 cp = w.w_space; 818 w.w_rtm.rtm_msglen = sizeof (struct rt_msghdr) + 819 2 * ROUNDUP_LONG(sizeof (struct sockaddr_in)); 820 w.w_rtm.rtm_version = RTM_VERSION; 821 w.w_rtm.rtm_type = action; 822 w.w_rtm.rtm_flags = flags; 823 w.w_rtm.rtm_seq = ++rt_sock_seqno; 824 w.w_rtm.rtm_addrs = RTA_DST|RTA_GATEWAY; 825 if (metric != 0 || action == RTM_CHANGE) { 826 w.w_rtm.rtm_rmx.rmx_hopcount = metric; 827 w.w_rtm.rtm_inits |= RTV_HOPCOUNT; 828 } 829 w.w_dst.sin_family = AF_INET; 830 w.w_dst.sin_addr.s_addr = dst; 831 w.w_gate.sin_family = AF_INET; 832 w.w_gate.sin_addr.s_addr = gate; 833 if (mask == HOST_MASK) { 834 w.w_rtm.rtm_flags |= RTF_HOST; 835 } else { 836 w.w_rtm.rtm_addrs |= RTA_NETMASK; 837 w_mask.sin_family = AF_INET; 838 w_mask.sin_addr.s_addr = htonl(mask); 839 (void) memmove(cp, &w_mask, sizeof (w_mask)); 840 cp += ROUNDUP_LONG(sizeof (struct sockaddr_in)); 841 w.w_rtm.rtm_msglen += ROUNDUP_LONG(sizeof (struct sockaddr_in)); 842 } 843 if (ifp == NULL) 844 ifp = iflookup(gate); 845 846 if ((ifp == NULL) || (ifp->int_phys == NULL)) { 847 trace_misc("no ifp for" PAT, ARGS); 848 } else { 849 if (ifp->int_phys->phyi_index > UINT16_MAX) { 850 trace_misc("ifindex %d is too big for sdl_index", 851 ifp->int_phys->phyi_index); 852 } else { 853 w_ifp.sdl_family = AF_LINK; 854 w.w_rtm.rtm_addrs |= RTA_IFP; 855 w_ifp.sdl_index = ifp->int_phys->phyi_index; 856 (void) memmove(cp, &w_ifp, sizeof (w_ifp)); 857 w.w_rtm.rtm_msglen += 858 ROUNDUP_LONG(sizeof (struct sockaddr_dl)); 859 } 860 } 861 862 863 if (!no_install) { 864 if (TRACERTS) 865 dump_rt_msg("write", &w.w_rtm, w.w_rtm.rtm_msglen); 866 cc = write(rt_sock, &w, w.w_rtm.rtm_msglen); 867 if (cc < 0) { 868 if (errno == ESRCH && (action == RTM_CHANGE || 869 action == RTM_DELETE)) { 870 trace_act("route disappeared before" PAT, ARGS); 871 if (action == RTM_CHANGE) { 872 action = RTM_ADD; 873 goto again; 874 } 875 return; 876 } 877 writelog(LOG_WARNING, "write(rt_sock)" PAT ": %s ", 878 ARGS, rip_strerror(errno)); 879 return; 880 } else if (cc != w.w_rtm.rtm_msglen) { 881 msglog("write(rt_sock) wrote %ld instead of %d for" PAT, 882 cc, w.w_rtm.rtm_msglen, ARGS); 883 return; 884 } 885 } 886 if (TRACEKERNEL) 887 trace_misc("write kernel" PAT, ARGS); 888 #undef PAT 889 #undef ARGS 890 } 891 892 893 /* Hash table containing our image of the kernel forwarding table. */ 894 #define KHASH_SIZE 71 /* should be prime */ 895 #define KHASH(a, m) khash_bins[((a) ^ (m)) % KHASH_SIZE] 896 static struct khash *khash_bins[KHASH_SIZE]; 897 898 #define K_KEEP_LIM 30 /* k_keep */ 899 900 static struct khash * 901 kern_find(in_addr_t dst, in_addr_t mask, in_addr_t gate, 902 struct interface *ifp, struct khash ***ppk) 903 { 904 struct khash *k, **pk; 905 906 if (ifp != NULL && ifp->int_phys != NULL) { 907 ifp = ifwithname(ifp->int_phys->phyi_name); 908 } 909 910 for (pk = &KHASH(dst, mask); (k = *pk) != NULL; pk = &k->k_next) { 911 if (k->k_dst == dst && k->k_mask == mask && 912 (gate == 0 || k->k_gate == gate) && 913 (ifp == NULL || k->k_ifp == ifp)) { 914 break; 915 } 916 } 917 if (ppk != NULL) 918 *ppk = pk; 919 return (k); 920 } 921 922 923 /* 924 * Find out if there is an alternate route to a given destination 925 * off of a given interface. 926 */ 927 static struct khash * 928 kern_alternate(in_addr_t dst, in_addr_t mask, in_addr_t gate, 929 struct interface *ifp, struct khash ***ppk) 930 { 931 struct khash *k, **pk; 932 933 if (ifp != NULL && ifp->int_phys != NULL) { 934 ifp = ifwithname(ifp->int_phys->phyi_name); 935 } 936 for (pk = &KHASH(dst, mask); (k = *pk) != NULL; pk = &k->k_next) { 937 if (k->k_dst == dst && k->k_mask == mask && 938 (k->k_gate != gate) && 939 (k->k_ifp == ifp)) { 940 break; 941 } 942 } 943 if (ppk != NULL) 944 *ppk = pk; 945 return (k); 946 } 947 948 static struct khash * 949 kern_add(in_addr_t dst, uint32_t mask, in_addr_t gate, struct interface *ifp) 950 { 951 struct khash *k, **pk; 952 953 if (ifp != NULL && ifp->int_phys != NULL) { 954 ifp = ifwithname(ifp->int_phys->phyi_name); 955 } 956 k = kern_find(dst, mask, gate, ifp, &pk); 957 if (k != NULL) 958 return (k); 959 960 k = rtmalloc(sizeof (*k), "kern_add"); 961 962 (void) memset(k, 0, sizeof (*k)); 963 k->k_dst = dst; 964 k->k_mask = mask; 965 k->k_state = KS_NEW; 966 k->k_keep = now.tv_sec; 967 k->k_gate = gate; 968 k->k_ifp = ifp; 969 *pk = k; 970 971 return (k); 972 } 973 974 /* delete all khash entries that are wired through the interface ifp */ 975 void 976 kern_flush_ifp(struct interface *ifp) 977 { 978 struct khash *k, *kprev, *knext; 979 int i; 980 981 if (ifp != NULL && ifp->int_phys != NULL) { 982 /* 983 * Only calculate phy ifp when the passed ifp is 984 * a logical IP interface. Otherwise the call 985 * ifwithname(phy ifname) will return NULL as we 986 * unlinked ifp from hashtables prior to this call 987 * in ifdel. 988 */ 989 if (strchr(ifp->int_phys->phyi_name, ':')) { 990 ifp = ifwithname(ifp->int_phys->phyi_name); 991 } 992 } 993 994 for (i = 0; i < KHASH_SIZE; i++) { 995 kprev = NULL; 996 for (k = khash_bins[i]; k != NULL; k = knext) { 997 knext = k->k_next; 998 if (k->k_ifp == ifp) { 999 if (kprev != NULL) 1000 kprev->k_next = k->k_next; 1001 else 1002 khash_bins[i] = k->k_next; 1003 free(k); 1004 continue; 1005 } 1006 kprev = k; 1007 } 1008 } 1009 } 1010 1011 /* 1012 * rewire khash entries that currently go through oldifp to 1013 * go through newifp. 1014 */ 1015 void 1016 kern_rewire_ifp(struct interface *oldifp, struct interface *newifp) 1017 { 1018 struct khash *k; 1019 int i; 1020 1021 if (oldifp != NULL && oldifp->int_phys != NULL) { 1022 oldifp = ifwithname(oldifp->int_phys->phyi_name); 1023 } 1024 if (newifp != NULL && newifp->int_phys != NULL) { 1025 newifp = ifwithname(newifp->int_phys->phyi_name); 1026 } 1027 for (i = 0; i < KHASH_SIZE; i++) { 1028 for (k = khash_bins[i]; k; k = k->k_next) { 1029 if (k->k_ifp == oldifp) { 1030 k->k_ifp = newifp; 1031 trace_misc("kern_rewire_ifp k 0x%lx " 1032 "from %s to %s", k, oldifp->int_name, 1033 newifp->int_name); 1034 } 1035 } 1036 } 1037 } 1038 1039 1040 /* 1041 * Check that a static route it is still in the daemon table, and not 1042 * deleted by interfaces coming and going. This is also the routine 1043 * responsible for adding new static routes to the daemon table. 1044 */ 1045 static void 1046 kern_check_static(struct khash *k, struct interface *ifp) 1047 { 1048 struct rt_entry *rt; 1049 struct rt_spare new; 1050 uint16_t rt_state = RS_STATIC; 1051 1052 if (ifp != NULL && ifp->int_phys != NULL) { 1053 ifp = ifwithname(ifp->int_phys->phyi_name); 1054 } 1055 (void) memset(&new, 0, sizeof (new)); 1056 new.rts_ifp = ifp; 1057 new.rts_gate = k->k_gate; 1058 new.rts_router = (ifp != NULL) ? ifp->int_addr : loopaddr; 1059 new.rts_metric = k->k_metric; 1060 new.rts_time = now.tv_sec; 1061 new.rts_origin = RO_STATIC; 1062 1063 rt = rtget(k->k_dst, k->k_mask); 1064 if ((ifp != NULL && !IS_IFF_ROUTING(ifp->int_if_flags)) || 1065 (k->k_state & KS_PRIVATE)) 1066 rt_state |= RS_NOPROPAGATE; 1067 1068 if (rt != NULL) { 1069 if ((rt->rt_state & RS_STATIC) == 0) { 1070 /* 1071 * We are already tracking this dest/mask 1072 * via RIP/RDISC. Ignore the static route, 1073 * because we don't currently have a good 1074 * way to compare metrics on static routes 1075 * with rip metrics, and therefore cannot 1076 * mix and match the two. 1077 */ 1078 return; 1079 } 1080 rt_state |= rt->rt_state; 1081 if (rt->rt_state != rt_state) 1082 rtchange(rt, rt_state, &new, 0); 1083 } else { 1084 rtadd(k->k_dst, k->k_mask, rt_state, &new); 1085 } 1086 } 1087 1088 1089 /* operate on a kernel entry */ 1090 static void 1091 kern_ioctl(struct khash *k, 1092 int action, /* RTM_DELETE, etc */ 1093 int flags) 1094 { 1095 if (((k->k_state & (KS_IF|KS_PASSIVE)) == KS_IF) || 1096 (k->k_state & KS_DEPRE_IF)) { 1097 /* 1098 * Prevent execution of RTM_DELETE, RTM_ADD or 1099 * RTM_CHANGE of interface routes 1100 */ 1101 trace_act("Blocking execution of %s %s --> %s ", 1102 rtm_type_name(action), 1103 addrname(k->k_dst, k->k_mask, 0), naddr_ntoa(k->k_gate)); 1104 return; 1105 } 1106 1107 switch (action) { 1108 case RTM_DELETE: 1109 k->k_state &= ~KS_DYNAMIC; 1110 if (k->k_state & KS_DELETED) 1111 return; 1112 k->k_state |= KS_DELETED; 1113 break; 1114 case RTM_ADD: 1115 k->k_state &= ~KS_DELETED; 1116 break; 1117 case RTM_CHANGE: 1118 if (k->k_state & KS_DELETED) { 1119 action = RTM_ADD; 1120 k->k_state &= ~KS_DELETED; 1121 } 1122 break; 1123 } 1124 1125 rtioctl(action, k->k_dst, k->k_gate, k->k_mask, k->k_ifp, 1126 k->k_metric, flags); 1127 } 1128 1129 1130 /* add a route the kernel told us */ 1131 static void 1132 rtm_add(struct rt_msghdr *rtm, 1133 struct rt_addrinfo *info, 1134 time_t keep, 1135 boolean_t interf_route, 1136 struct interface *ifptr) 1137 { 1138 struct khash *k; 1139 struct interface *ifp = ifptr; 1140 in_addr_t mask, gate = 0; 1141 static struct msg_limit msg_no_ifp; 1142 1143 if (rtm->rtm_flags & RTF_HOST) { 1144 mask = HOST_MASK; 1145 } else if (INFO_MASK(info) != 0) { 1146 mask = ntohl(S_ADDR(INFO_MASK(info))); 1147 } else { 1148 writelog(LOG_WARNING, 1149 "ignore %s without mask", rtm_type_name(rtm->rtm_type)); 1150 return; 1151 } 1152 1153 /* 1154 * Find the interface toward the gateway. 1155 */ 1156 if (INFO_GATE(info) != NULL) 1157 gate = S_ADDR(INFO_GATE(info)); 1158 1159 if (ifp == NULL) { 1160 if (INFO_GATE(info) != NULL) 1161 ifp = iflookup(gate); 1162 if (ifp == NULL) { 1163 msglim(&msg_no_ifp, gate, 1164 "route %s --> %s nexthop is not directly connected", 1165 addrname(S_ADDR(INFO_DST(info)), mask, 0), 1166 naddr_ntoa(gate)); 1167 } else { 1168 if (ifp->int_phys != NULL) { 1169 ifp = ifwithname(ifp->int_phys->phyi_name); 1170 } 1171 } 1172 } 1173 1174 k = kern_add(S_ADDR(INFO_DST(info)), mask, gate, ifp); 1175 1176 if (k->k_state & KS_NEW) 1177 k->k_keep = now.tv_sec+keep; 1178 if (INFO_GATE(info) == 0) { 1179 trace_act("note %s without gateway", 1180 rtm_type_name(rtm->rtm_type)); 1181 k->k_metric = HOPCNT_INFINITY; 1182 } else if (INFO_GATE(info)->ss_family != AF_INET) { 1183 trace_act("note %s with gateway AF=%d", 1184 rtm_type_name(rtm->rtm_type), 1185 INFO_GATE(info)->ss_family); 1186 k->k_metric = HOPCNT_INFINITY; 1187 } else { 1188 k->k_gate = S_ADDR(INFO_GATE(info)); 1189 k->k_metric = rtm->rtm_rmx.rmx_hopcount; 1190 if (k->k_metric < 0) 1191 k->k_metric = 0; 1192 else if (k->k_metric > HOPCNT_INFINITY-1) 1193 k->k_metric = HOPCNT_INFINITY-1; 1194 } 1195 1196 if ((k->k_state & KS_NEW) && interf_route) { 1197 if (k->k_gate != 0 && findifaddr(k->k_gate) == NULL) 1198 k->k_state |= KS_DEPRE_IF; 1199 else 1200 k->k_state |= KS_IF; 1201 } 1202 1203 k->k_state &= ~(KS_NEW | KS_DELETE | KS_ADD | KS_CHANGE | KS_DEL_ADD | 1204 KS_STATIC | KS_GATEWAY | KS_DELETED | KS_PRIVATE | KS_CHECK); 1205 if (rtm->rtm_flags & RTF_GATEWAY) 1206 k->k_state |= KS_GATEWAY; 1207 if (rtm->rtm_flags & RTF_STATIC) 1208 k->k_state |= KS_STATIC; 1209 if (rtm->rtm_flags & RTF_PRIVATE) 1210 k->k_state |= KS_PRIVATE; 1211 1212 1213 if (rtm->rtm_flags & (RTF_DYNAMIC | RTF_MODIFIED)) { 1214 if (INFO_AUTHOR(info) != 0 && 1215 INFO_AUTHOR(info)->ss_family == AF_INET) 1216 ifp = iflookup(S_ADDR(INFO_AUTHOR(info))); 1217 else 1218 ifp = NULL; 1219 if (should_supply(ifp) && (ifp == NULL || 1220 !(ifp->int_state & IS_REDIRECT_OK))) { 1221 /* 1222 * Routers are not supposed to listen to redirects, 1223 * so delete it if it came via an unknown interface 1224 * or the interface does not have special permission. 1225 */ 1226 k->k_state &= ~KS_DYNAMIC; 1227 k->k_state |= KS_DELETE; 1228 LIM_SEC(need_kern, 0); 1229 trace_act("mark for deletion redirected %s --> %s" 1230 " via %s", 1231 addrname(k->k_dst, k->k_mask, 0), 1232 naddr_ntoa(k->k_gate), 1233 ifp ? ifp->int_name : "unknown interface"); 1234 } else { 1235 k->k_state |= KS_DYNAMIC; 1236 k->k_redirect_time = now.tv_sec; 1237 trace_act("accept redirected %s --> %s via %s", 1238 addrname(k->k_dst, k->k_mask, 0), 1239 naddr_ntoa(k->k_gate), 1240 ifp ? ifp->int_name : "unknown interface"); 1241 } 1242 return; 1243 } 1244 1245 /* 1246 * If it is not a static route, quit until the next comparison 1247 * between the kernel and daemon tables, when it will be deleted. 1248 */ 1249 if (!(k->k_state & KS_STATIC)) { 1250 if (!(k->k_state & (KS_IF|KS_DEPRE_IF|KS_FILE))) 1251 k->k_state |= KS_DELETE; 1252 LIM_SEC(need_kern, k->k_keep); 1253 return; 1254 } 1255 1256 /* 1257 * Put static routes with real metrics into the daemon table so 1258 * they can be advertised. 1259 */ 1260 1261 kern_check_static(k, ifp); 1262 } 1263 1264 1265 /* deal with packet loss */ 1266 static void 1267 rtm_lose(struct rt_msghdr *rtm, struct rt_addrinfo *info) 1268 { 1269 if (INFO_GATE(info) == NULL || INFO_GATE(info)->ss_family != AF_INET) { 1270 trace_act("ignore %s without gateway", 1271 rtm_type_name(rtm->rtm_type)); 1272 age(0); 1273 return; 1274 } 1275 1276 if (rdisc_ok) 1277 rdisc_age(S_ADDR(INFO_GATE(info))); 1278 age(S_ADDR(INFO_GATE(info))); 1279 } 1280 1281 1282 /* 1283 * Make the gateway slot of an info structure point to something 1284 * useful. If it is not already useful, but it specifies an interface, 1285 * then fill in the sockaddr_in provided and point it there. 1286 */ 1287 static int 1288 get_info_gate(struct sockaddr_storage **ssp, struct sockaddr_in *sin) 1289 { 1290 struct sockaddr_dl *sdl = (struct sockaddr_dl *)*ssp; 1291 struct interface *ifp; 1292 1293 if (sdl == NULL) 1294 return (0); 1295 if ((sdl)->sdl_family == AF_INET) 1296 return (1); 1297 if ((sdl)->sdl_family != AF_LINK) 1298 return (0); 1299 1300 ifp = ifwithindex(sdl->sdl_index, _B_TRUE); 1301 if (ifp == NULL) 1302 return (0); 1303 1304 sin->sin_addr.s_addr = ifp->int_addr; 1305 sin->sin_family = AF_INET; 1306 /* LINTED */ 1307 *ssp = (struct sockaddr_storage *)sin; 1308 1309 return (1); 1310 } 1311 1312 1313 /* 1314 * Clean the kernel table by copying it to the daemon image. 1315 * Eventually the daemon will delete any extra routes. 1316 */ 1317 void 1318 sync_kern(void) 1319 { 1320 int i; 1321 struct khash *k; 1322 struct { 1323 struct T_optmgmt_req req; 1324 struct opthdr hdr; 1325 } req; 1326 union { 1327 struct T_optmgmt_ack ack; 1328 unsigned char space[64]; 1329 } ack; 1330 struct opthdr *rh; 1331 struct strbuf cbuf, dbuf; 1332 int ipfd, nroutes, flags, r; 1333 mib2_ipRouteEntry_t routes[8]; 1334 mib2_ipRouteEntry_t *rp; 1335 struct rt_msghdr rtm; 1336 struct rt_addrinfo info; 1337 struct sockaddr_in sin_dst; 1338 struct sockaddr_in sin_gate; 1339 struct sockaddr_in sin_mask; 1340 struct sockaddr_in sin_author; 1341 struct interface *ifp; 1342 char ifname[LIFNAMSIZ + 1]; 1343 1344 for (i = 0; i < KHASH_SIZE; i++) { 1345 for (k = khash_bins[i]; k != NULL; k = k->k_next) { 1346 if (!(k->k_state & (KS_IF|KS_DEPRE_IF))) 1347 k->k_state |= KS_CHECK; 1348 } 1349 } 1350 1351 ipfd = open(IP_DEV_NAME, O_RDWR); 1352 if (ipfd == -1) { 1353 msglog("open " IP_DEV_NAME ": %s", rip_strerror(errno)); 1354 goto hash_clean; 1355 } 1356 1357 req.req.PRIM_type = T_OPTMGMT_REQ; 1358 req.req.OPT_offset = (caddr_t)&req.hdr - (caddr_t)&req; 1359 req.req.OPT_length = sizeof (req.hdr); 1360 req.req.MGMT_flags = T_CURRENT; 1361 1362 req.hdr.level = MIB2_IP; 1363 req.hdr.name = 0; 1364 req.hdr.len = 0; 1365 1366 cbuf.buf = (caddr_t)&req; 1367 cbuf.len = sizeof (req); 1368 1369 if (putmsg(ipfd, &cbuf, NULL, 0) == -1) { 1370 msglog("T_OPTMGMT_REQ putmsg: %s", rip_strerror(errno)); 1371 goto hash_clean; 1372 } 1373 1374 for (;;) { 1375 cbuf.buf = (caddr_t)&ack; 1376 cbuf.maxlen = sizeof (ack); 1377 dbuf.buf = (caddr_t)routes; 1378 dbuf.maxlen = sizeof (routes); 1379 flags = 0; 1380 r = getmsg(ipfd, &cbuf, &dbuf, &flags); 1381 if (r == -1) { 1382 msglog("T_OPTMGMT_REQ getmsg: %s", rip_strerror(errno)); 1383 goto hash_clean; 1384 } 1385 1386 if (cbuf.len < sizeof (struct T_optmgmt_ack) || 1387 ack.ack.PRIM_type != T_OPTMGMT_ACK || 1388 ack.ack.MGMT_flags != T_SUCCESS || 1389 ack.ack.OPT_length < sizeof (struct opthdr)) { 1390 msglog("bad T_OPTMGMT response; len=%d prim=%d " 1391 "flags=%d optlen=%d", cbuf.len, ack.ack.PRIM_type, 1392 ack.ack.MGMT_flags, ack.ack.OPT_length); 1393 goto hash_clean; 1394 } 1395 /* LINTED */ 1396 rh = (struct opthdr *)((caddr_t)&ack + ack.ack.OPT_offset); 1397 if (rh->level == 0 && rh->name == 0) { 1398 break; 1399 } 1400 if (rh->level != MIB2_IP || rh->name != MIB2_IP_21) { 1401 while (r == MOREDATA) { 1402 r = getmsg(ipfd, NULL, &dbuf, &flags); 1403 } 1404 continue; 1405 } 1406 break; 1407 } 1408 1409 (void) memset(&rtm, 0, sizeof (rtm)); 1410 (void) memset(&info, 0, sizeof (info)); 1411 (void) memset(&sin_dst, 0, sizeof (sin_dst)); 1412 (void) memset(&sin_gate, 0, sizeof (sin_gate)); 1413 (void) memset(&sin_mask, 0, sizeof (sin_mask)); 1414 (void) memset(&sin_author, 0, sizeof (sin_author)); 1415 sin_dst.sin_family = AF_INET; 1416 /* LINTED */ 1417 info.rti_info[RTAX_DST] = (struct sockaddr_storage *)&sin_dst; 1418 sin_gate.sin_family = AF_INET; 1419 /* LINTED */ 1420 info.rti_info[RTAX_GATEWAY] = (struct sockaddr_storage *)&sin_gate; 1421 sin_mask.sin_family = AF_INET; 1422 /* LINTED */ 1423 info.rti_info[RTAX_NETMASK] = (struct sockaddr_storage *)&sin_mask; 1424 sin_dst.sin_family = AF_INET; 1425 /* LINTED */ 1426 info.rti_info[RTAX_AUTHOR] = (struct sockaddr_storage *)&sin_author; 1427 1428 for (;;) { 1429 nroutes = dbuf.len / sizeof (mib2_ipRouteEntry_t); 1430 for (rp = routes; nroutes > 0; ++rp, nroutes--) { 1431 1432 /* 1433 * Ignore IRE cache, broadcast, and local address 1434 * entries; they're not subject to routing socket 1435 * control. 1436 */ 1437 if (rp->ipRouteInfo.re_ire_type & 1438 (IRE_BROADCAST | IRE_CACHE | IRE_LOCAL)) 1439 continue; 1440 1441 /* ignore multicast and link local addresses */ 1442 if (IN_MULTICAST(ntohl(rp->ipRouteDest)) || 1443 IN_LINKLOCAL(ntohl(rp->ipRouteDest))) { 1444 continue; 1445 } 1446 1447 1448 #ifdef DEBUG_KERNEL_ROUTE_READ 1449 (void) fprintf(stderr, "route type %d, ire type %08X, " 1450 "flags %08X: %s", rp->ipRouteType, 1451 rp->ipRouteInfo.re_ire_type, 1452 rp->ipRouteInfo.re_flags, 1453 naddr_ntoa(rp->ipRouteDest)); 1454 (void) fprintf(stderr, " %s", 1455 naddr_ntoa(rp->ipRouteMask)); 1456 (void) fprintf(stderr, " %s\n", 1457 naddr_ntoa(rp->ipRouteNextHop)); 1458 #endif 1459 1460 /* Fake up the needed entries */ 1461 rtm.rtm_flags = rp->ipRouteInfo.re_flags; 1462 rtm.rtm_type = RTM_GET; 1463 rtm.rtm_rmx.rmx_hopcount = rp->ipRouteMetric1; 1464 1465 (void) memset(ifname, 0, sizeof (ifname)); 1466 if (rp->ipRouteIfIndex.o_length < 1467 sizeof (rp->ipRouteIfIndex.o_bytes)) 1468 rp->ipRouteIfIndex.o_bytes[ 1469 rp->ipRouteIfIndex.o_length] = '\0'; 1470 (void) strncpy(ifname, 1471 rp->ipRouteIfIndex.o_bytes, 1472 sizeof (ifname)); 1473 1474 /* 1475 * First try to match up on gwkludge entries 1476 * before trying to match ifp by name. 1477 */ 1478 if ((ifp = gwkludge_iflookup(rp->ipRouteDest, 1479 rp->ipRouteNextHop, 1480 ntohl(rp->ipRouteMask))) == NULL) { 1481 ifp = ifwithname(ifname); 1482 if (ifp != NULL && ifp->int_phys != NULL) { 1483 ifp = ifwithname( 1484 ifp->int_phys->phyi_name); 1485 } 1486 } 1487 1488 info.rti_addrs = RTA_DST | RTA_GATEWAY | RTA_NETMASK; 1489 if (rp->ipRouteInfo.re_ire_type & IRE_HOST_REDIRECT) 1490 info.rti_addrs |= RTA_AUTHOR; 1491 sin_dst.sin_addr.s_addr = rp->ipRouteDest; 1492 sin_gate.sin_addr.s_addr = rp->ipRouteNextHop; 1493 sin_mask.sin_addr.s_addr = rp->ipRouteMask; 1494 sin_author.sin_addr.s_addr = 1495 rp->ipRouteInfo.re_src_addr; 1496 1497 /* 1498 * Note static routes and interface routes, and also 1499 * preload the image of the kernel table so that 1500 * we can later clean it, as well as avoid making 1501 * unneeded changes. Keep the old kernel routes for a 1502 * few seconds to allow a RIP or router-discovery 1503 * response to be heard. 1504 */ 1505 rtm_add(&rtm, &info, MAX_WAITTIME, 1506 ((rp->ipRouteInfo.re_ire_type & 1507 (IRE_INTERFACE|IRE_LOOPBACK)) != 0), ifp); 1508 } 1509 if (r == 0) { 1510 break; 1511 } 1512 r = getmsg(ipfd, NULL, &dbuf, &flags); 1513 } 1514 1515 hash_clean: 1516 if (ipfd != -1) 1517 (void) close(ipfd); 1518 for (i = 0; i < KHASH_SIZE; i++) { 1519 for (k = khash_bins[i]; k != NULL; k = k->k_next) { 1520 1521 /* 1522 * KS_DELETED routes have been removed from the 1523 * kernel, but we keep them around for reasons 1524 * stated in del_static(), so we skip the check 1525 * for KS_DELETED routes here. 1526 */ 1527 if ((k->k_state & (KS_CHECK|KS_DELETED)) == KS_CHECK) { 1528 1529 if (!(k->k_state & KS_DYNAMIC)) { 1530 writelog(LOG_WARNING, 1531 "%s --> %s disappeared from kernel", 1532 addrname(k->k_dst, k->k_mask, 0), 1533 naddr_ntoa(k->k_gate)); 1534 } 1535 del_static(k->k_dst, k->k_mask, k->k_gate, 1536 k->k_ifp, 1); 1537 1538 } 1539 } 1540 } 1541 } 1542 1543 1544 /* Listen to announcements from the kernel */ 1545 void 1546 read_rt(void) 1547 { 1548 long cc; 1549 struct interface *ifp; 1550 struct sockaddr_in gate_sin; 1551 in_addr_t mask, gate; 1552 union { 1553 struct { 1554 struct rt_msghdr rtm; 1555 struct sockaddr_storage addrs[RTA_NUMBITS]; 1556 } r; 1557 struct if_msghdr ifm; 1558 } m; 1559 char str[100], *strp; 1560 struct rt_addrinfo info; 1561 1562 1563 for (;;) { 1564 cc = read(rt_sock, &m, sizeof (m)); 1565 if (cc <= 0) { 1566 if (cc < 0 && errno != EWOULDBLOCK) 1567 LOGERR("read(rt_sock)"); 1568 return; 1569 } 1570 1571 if (TRACERTS) 1572 dump_rt_msg("read", &m.r.rtm, cc); 1573 1574 if (cc < m.r.rtm.rtm_msglen) { 1575 msglog("routing message truncated (%d < %d)", 1576 cc, m.r.rtm.rtm_msglen); 1577 } 1578 1579 if (m.r.rtm.rtm_version != RTM_VERSION) { 1580 msglog("bogus routing message version %d", 1581 m.r.rtm.rtm_version); 1582 continue; 1583 } 1584 1585 ifp = NULL; 1586 1587 if (m.r.rtm.rtm_type == RTM_IFINFO || 1588 m.r.rtm.rtm_type == RTM_NEWADDR || 1589 m.r.rtm.rtm_type == RTM_DELADDR) { 1590 strp = if_bit_string(m.ifm.ifm_flags, _B_TRUE); 1591 if (strp == NULL) { 1592 strp = str; 1593 (void) sprintf(str, "%#x", m.ifm.ifm_flags); 1594 } 1595 ifp = ifwithindex(m.ifm.ifm_index, 1596 m.r.rtm.rtm_type != RTM_DELADDR); 1597 if (ifp == NULL) { 1598 char ifname[LIFNAMSIZ], *ifnamep; 1599 1600 ifnamep = if_indextoname(m.ifm.ifm_index, 1601 ifname); 1602 if (ifnamep == NULL) { 1603 trace_act("note %s with flags %s" 1604 " for unknown interface index #%d", 1605 rtm_type_name(m.r.rtm.rtm_type), 1606 strp, m.ifm.ifm_index); 1607 } else { 1608 trace_act("note %s with flags %s" 1609 " for unknown interface %s", 1610 rtm_type_name(m.r.rtm.rtm_type), 1611 strp, ifnamep); 1612 } 1613 } else { 1614 trace_act("note %s with flags %s for %s", 1615 rtm_type_name(m.r.rtm.rtm_type), 1616 strp, ifp->int_name); 1617 } 1618 if (strp != str) 1619 free(strp); 1620 1621 /* 1622 * After being informed of a change to an interface, 1623 * check them all now if the check would otherwise 1624 * be a long time from now, if the interface is 1625 * not known, or if the interface has been turned 1626 * off or on. 1627 */ 1628 if (ifscan_timer.tv_sec-now.tv_sec >= 1629 CHECK_BAD_INTERVAL || ifp == NULL || 1630 ((ifp->int_if_flags ^ m.ifm.ifm_flags) & 1631 IFF_UP) != 0) 1632 ifscan_timer.tv_sec = now.tv_sec; 1633 continue; 1634 } else { 1635 if (m.r.rtm.rtm_index != 0) 1636 ifp = ifwithindex(m.r.rtm.rtm_index, 1); 1637 } 1638 1639 (void) strlcpy(str, rtm_type_name(m.r.rtm.rtm_type), 1640 sizeof (str)); 1641 strp = &str[strlen(str)]; 1642 if (m.r.rtm.rtm_type <= RTM_CHANGE) 1643 strp += snprintf(strp, sizeof (str) - (strp - str), 1644 " from pid %d", (int)m.r.rtm.rtm_pid); 1645 1646 /* LINTED */ 1647 (void) rt_xaddrs(&info, (struct sockaddr_storage *)(&m.r.rtm + 1648 1), (char *)&m + cc, m.r.rtm.rtm_addrs); 1649 1650 if (INFO_DST(&info) == 0) { 1651 trace_act("ignore %s without dst", str); 1652 continue; 1653 } 1654 1655 if (INFO_DST(&info)->ss_family != AF_INET) { 1656 trace_act("ignore %s for AF %d", str, 1657 INFO_DST(&info)->ss_family); 1658 continue; 1659 } 1660 1661 mask = ((INFO_MASK(&info) != 0) ? 1662 ntohl(S_ADDR(INFO_MASK(&info))) : 1663 (m.r.rtm.rtm_flags & RTF_HOST) ? 1664 HOST_MASK : std_mask(S_ADDR(INFO_DST(&info)))); 1665 1666 strp += snprintf(strp, sizeof (str) - (strp - str), ": %s", 1667 addrname(S_ADDR(INFO_DST(&info)), mask, 0)); 1668 1669 if (IN_MULTICAST(ntohl(S_ADDR(INFO_DST(&info)))) || 1670 IN_LINKLOCAL(ntohl(S_ADDR(INFO_DST(&info))))) { 1671 trace_act("ignore multicast/link local %s", str); 1672 continue; 1673 } 1674 1675 if (m.r.rtm.rtm_flags & RTF_LLINFO) { 1676 trace_act("ignore ARP %s", str); 1677 continue; 1678 } 1679 1680 if (get_info_gate(&INFO_GATE(&info), &gate_sin)) { 1681 gate = S_ADDR(INFO_GATE(&info)); 1682 strp += snprintf(strp, sizeof (str) - (strp - str), 1683 " --> %s", naddr_ntoa(gate)); 1684 } else { 1685 gate = 0; 1686 } 1687 1688 if (INFO_AUTHOR(&info) != 0) 1689 strp += snprintf(strp, sizeof (str) - (strp - str), 1690 " by authority of %s", 1691 saddr_ntoa(INFO_AUTHOR(&info))); 1692 1693 switch (m.r.rtm.rtm_type) { 1694 case RTM_ADD: 1695 case RTM_CHANGE: 1696 case RTM_REDIRECT: 1697 if (m.r.rtm.rtm_errno != 0) { 1698 trace_act("ignore %s with \"%s\" error", 1699 str, rip_strerror(m.r.rtm.rtm_errno)); 1700 } else { 1701 trace_act("%s", str); 1702 rtm_add(&m.r.rtm, &info, 0, 1703 !(m.r.rtm.rtm_flags & RTF_GATEWAY) && 1704 m.r.rtm.rtm_type != RTM_REDIRECT, ifp); 1705 1706 } 1707 break; 1708 1709 case RTM_DELETE: 1710 if (m.r.rtm.rtm_errno != 0 && 1711 m.r.rtm.rtm_errno != ESRCH) { 1712 trace_act("ignore %s with \"%s\" error", 1713 str, rip_strerror(m.r.rtm.rtm_errno)); 1714 } else { 1715 trace_act("%s", str); 1716 del_static(S_ADDR(INFO_DST(&info)), mask, 1717 gate, ifp, 1); 1718 } 1719 break; 1720 1721 case RTM_LOSING: 1722 trace_act("%s", str); 1723 rtm_lose(&m.r.rtm, &info); 1724 break; 1725 1726 default: 1727 trace_act("ignore %s", str); 1728 break; 1729 } 1730 } 1731 } 1732 1733 1734 /* 1735 * Disassemble a routing message. The result is an array of pointers 1736 * to sockaddr_storage structures stored in the info argument. 1737 * 1738 * ss is a pointer to the beginning of the data following the 1739 * rt_msghdr contained in the routing socket message, which consists 1740 * of a string of concatenated sockaddr structure of different types. 1741 * 1742 * Extended attributes can be appended at the end of the list. 1743 */ 1744 static int 1745 rt_xaddrs(struct rt_addrinfo *info, 1746 struct sockaddr_storage *ss, 1747 char *lim, 1748 int addrs) 1749 { 1750 int retv = 0; 1751 int i; 1752 int abit; 1753 int complaints; 1754 static int prev_complaints; 1755 1756 #define XBAD_AF 0x1 1757 #define XBAD_SHORT 0x2 1758 #define XBAD_LONG 0x4 1759 1760 (void) memset(info, 0, sizeof (*info)); 1761 info->rti_addrs = addrs; 1762 complaints = 0; 1763 for (i = 0, abit = 1; i < RTAX_MAX && (char *)ss < lim; 1764 i++, abit <<= 1) { 1765 if ((addrs & abit) == 0) 1766 continue; 1767 info->rti_info[i] = ss; 1768 /* Horrible interface here */ 1769 switch (ss->ss_family) { 1770 case AF_UNIX: 1771 /* LINTED */ 1772 ss = (struct sockaddr_storage *)( 1773 (struct sockaddr_un *)ss + 1); 1774 break; 1775 case AF_INET: 1776 /* LINTED */ 1777 ss = (struct sockaddr_storage *)( 1778 (struct sockaddr_in *)ss + 1); 1779 break; 1780 case AF_LINK: 1781 /* LINTED */ 1782 ss = (struct sockaddr_storage *)( 1783 (struct sockaddr_dl *)ss + 1); 1784 break; 1785 case AF_INET6: 1786 /* LINTED */ 1787 ss = (struct sockaddr_storage *)( 1788 (struct sockaddr_in6 *)ss + 1); 1789 break; 1790 default: 1791 if (!(prev_complaints & XBAD_AF)) 1792 writelog(LOG_WARNING, 1793 "unknown address family %d " 1794 "encountered", ss->ss_family); 1795 if (complaints & XBAD_AF) 1796 goto xaddr_done; 1797 /* LINTED */ 1798 ss = (struct sockaddr_storage *)( 1799 (struct sockaddr *)ss + 1); 1800 complaints |= XBAD_AF; 1801 info->rti_addrs &= abit - 1; 1802 addrs = info->rti_addrs; 1803 retv = -1; 1804 break; 1805 } 1806 if ((char *)ss > lim) { 1807 if (!(prev_complaints & XBAD_SHORT)) 1808 msglog("sockaddr %d too short by %d " 1809 "bytes", i + 1, (char *)ss - lim); 1810 complaints |= XBAD_SHORT; 1811 info->rti_info[i] = NULL; 1812 info->rti_addrs &= abit - 1; 1813 retv = -1; 1814 goto xaddr_done; 1815 } 1816 } 1817 1818 while (((char *)ss + sizeof (rtm_ext_t)) <= lim) { 1819 rtm_ext_t *tp; 1820 char *nxt; 1821 1822 /* LINTED: alignment */ 1823 tp = (rtm_ext_t *)ss; 1824 nxt = (char *)(tp + 1) + tp->rtmex_len; 1825 1826 if (!IS_P2ALIGNED(tp->rtmex_len, sizeof (uint32_t)) || 1827 nxt > lim) { 1828 break; 1829 } 1830 1831 /* LINTED: alignment */ 1832 ss = (struct sockaddr_storage *)nxt; 1833 } 1834 1835 if ((char *)ss != lim) { 1836 if ((char *)ss > lim) { 1837 if (!(prev_complaints & XBAD_SHORT)) 1838 msglog("routing message too short by %d bytes", 1839 (char *)ss - lim); 1840 complaints |= XBAD_SHORT; 1841 } else if (!(prev_complaints & XBAD_LONG)) { 1842 msglog("%d bytes of routing message left over", 1843 lim - (char *)ss); 1844 complaints |= XBAD_LONG; 1845 } 1846 retv = -1; 1847 } 1848 xaddr_done: 1849 prev_complaints = complaints; 1850 return (retv); 1851 } 1852 1853 1854 /* after aggregating, note routes that belong in the kernel */ 1855 static void 1856 kern_out(struct ag_info *ag) 1857 { 1858 struct khash *k; 1859 struct interface *ifp; 1860 1861 ifp = ag->ag_ifp; 1862 1863 if (ifp != NULL && ifp->int_phys != NULL) { 1864 ifp = ifwithname(ifp->int_phys->phyi_name); 1865 } 1866 1867 /* 1868 * Do not install bad routes if they are not already present. 1869 * This includes routes that had RS_NET_SYN for interfaces that 1870 * recently died. 1871 */ 1872 if (ag->ag_metric == HOPCNT_INFINITY) { 1873 k = kern_find(htonl(ag->ag_dst_h), ag->ag_mask, 1874 ag->ag_nhop, ag->ag_ifp, NULL); 1875 if (k == NULL) 1876 return; 1877 } else { 1878 k = kern_add(htonl(ag->ag_dst_h), ag->ag_mask, ag->ag_nhop, 1879 ifp); 1880 } 1881 1882 if (k->k_state & KS_NEW) { 1883 /* will need to add new entry to the kernel table */ 1884 k->k_state = KS_ADD; 1885 if (ag->ag_state & AGS_GATEWAY) 1886 k->k_state |= KS_GATEWAY; 1887 if (ag->ag_state & AGS_IF) 1888 k->k_state |= KS_IF; 1889 if (ag->ag_state & AGS_PASSIVE) 1890 k->k_state |= KS_PASSIVE; 1891 if (ag->ag_state & AGS_FILE) 1892 k->k_state |= KS_FILE; 1893 k->k_gate = ag->ag_nhop; 1894 k->k_ifp = ifp; 1895 k->k_metric = ag->ag_metric; 1896 return; 1897 } 1898 1899 if ((k->k_state & (KS_STATIC|KS_DEPRE_IF)) || 1900 ((k->k_state & (KS_IF|KS_PASSIVE)) == KS_IF)) { 1901 return; 1902 } 1903 1904 /* modify existing kernel entry if necessary */ 1905 if (k->k_gate == ag->ag_nhop && k->k_ifp == ag->ag_ifp && 1906 k->k_metric != ag->ag_metric) { 1907 /* 1908 * Must delete bad interface routes etc. 1909 * to change them. 1910 */ 1911 if (k->k_metric == HOPCNT_INFINITY) 1912 k->k_state |= KS_DEL_ADD; 1913 k->k_gate = ag->ag_nhop; 1914 k->k_metric = ag->ag_metric; 1915 k->k_state |= KS_CHANGE; 1916 } 1917 1918 /* 1919 * If the daemon thinks the route should exist, forget 1920 * about any redirections. 1921 * If the daemon thinks the route should exist, eventually 1922 * override manual intervention by the operator. 1923 */ 1924 if ((k->k_state & (KS_DYNAMIC | KS_DELETED)) != 0) { 1925 k->k_state &= ~KS_DYNAMIC; 1926 k->k_state |= (KS_ADD | KS_DEL_ADD); 1927 } 1928 1929 if ((k->k_state & KS_GATEWAY) && !(ag->ag_state & AGS_GATEWAY)) { 1930 k->k_state &= ~KS_GATEWAY; 1931 k->k_state |= (KS_ADD | KS_DEL_ADD); 1932 } else if (!(k->k_state & KS_GATEWAY) && (ag->ag_state & AGS_GATEWAY)) { 1933 k->k_state |= KS_GATEWAY; 1934 k->k_state |= (KS_ADD | KS_DEL_ADD); 1935 } 1936 1937 /* 1938 * Deleting-and-adding is necessary to change aspects of a route. 1939 * Just delete instead of deleting and then adding a bad route. 1940 * Otherwise, we want to keep the route in the kernel. 1941 */ 1942 if (k->k_metric == HOPCNT_INFINITY && (k->k_state & KS_DEL_ADD)) 1943 k->k_state |= KS_DELETE; 1944 else 1945 k->k_state &= ~KS_DELETE; 1946 #undef RT 1947 } 1948 1949 /* 1950 * Update our image of the kernel forwarding table using the given 1951 * route from our internal routing table. 1952 */ 1953 1954 /*ARGSUSED1*/ 1955 static int 1956 walk_kern(struct radix_node *rn, void *argp) 1957 { 1958 #define RT ((struct rt_entry *)rn) 1959 uint8_t metric, pref; 1960 uint_t ags = 0; 1961 int i; 1962 struct rt_spare *rts; 1963 1964 /* Do not install synthetic routes */ 1965 if (RT->rt_state & RS_NET_SYN) 1966 return (0); 1967 1968 /* 1969 * Do not install static routes here. Only 1970 * read_rt->rtm_add->kern_add should install those 1971 */ 1972 if ((RT->rt_state & RS_STATIC) && 1973 (RT->rt_spares[0].rts_origin != RO_FILE)) 1974 return (0); 1975 1976 /* Do not clobber kernel if this is a route for a dead interface */ 1977 if (RT->rt_state & RS_BADIF) 1978 return (0); 1979 1980 if (!(RT->rt_state & RS_IF)) { 1981 /* This is an ordinary route, not for an interface. */ 1982 1983 /* 1984 * aggregate, ordinary good routes without regard to 1985 * their metric 1986 */ 1987 pref = 1; 1988 ags |= (AGS_GATEWAY | AGS_SUPPRESS | AGS_AGGREGATE); 1989 1990 /* 1991 * Do not install host routes directly to hosts, to avoid 1992 * interfering with ARP entries in the kernel table. 1993 */ 1994 if (RT_ISHOST(RT) && ntohl(RT->rt_dst) == RT->rt_gate) 1995 return (0); 1996 1997 } else { 1998 /* 1999 * This is an interface route. 2000 * Do not install routes for "external" remote interfaces. 2001 */ 2002 if (RT->rt_ifp != NULL && (RT->rt_ifp->int_state & IS_EXTERNAL)) 2003 return (0); 2004 2005 /* Interfaces should override received routes. */ 2006 pref = 0; 2007 ags |= (AGS_IF | AGS_CORS_GATE); 2008 if (RT->rt_ifp != NULL && 2009 !(RT->rt_ifp->int_if_flags & IFF_LOOPBACK) && 2010 (RT->rt_ifp->int_state & (IS_PASSIVE|IS_ALIAS)) == 2011 IS_PASSIVE) { 2012 ags |= AGS_PASSIVE; 2013 } 2014 2015 /* 2016 * If it is not an interface, or an alias for an interface, 2017 * it must be a "gateway." 2018 * 2019 * If it is a "remote" interface, it is also a "gateway" to 2020 * the kernel if is not a alias. 2021 */ 2022 if (RT->rt_ifp == NULL || (RT->rt_ifp->int_state & IS_REMOTE)) { 2023 2024 ags |= (AGS_GATEWAY | AGS_SUPPRESS); 2025 2026 /* 2027 * Do not aggregate IS_PASSIVE routes. 2028 */ 2029 if (!(RT->rt_ifp->int_state & IS_PASSIVE)) 2030 ags |= AGS_AGGREGATE; 2031 } 2032 } 2033 2034 metric = RT->rt_metric; 2035 if (metric == HOPCNT_INFINITY) { 2036 /* If the route is dead, try hard to aggregate. */ 2037 pref = HOPCNT_INFINITY; 2038 ags |= (AGS_FINE_GATE | AGS_SUPPRESS); 2039 ags &= ~(AGS_IF | AGS_CORS_GATE); 2040 } 2041 2042 /* 2043 * dump all routes that have the same metric as rt_spares[0] 2044 * into the kern_table, to be added to the kernel. 2045 */ 2046 for (i = 0; i < RT->rt_num_spares; i++) { 2047 rts = &RT->rt_spares[i]; 2048 2049 /* Do not install external routes */ 2050 if (rts->rts_flags & RTS_EXTERNAL) 2051 continue; 2052 2053 if (rts->rts_metric == metric) { 2054 ag_check(RT->rt_dst, RT->rt_mask, 2055 rts->rts_router, rts->rts_ifp, rts->rts_gate, 2056 metric, pref, 0, 0, 2057 (rts->rts_origin & RO_FILE) ? (ags|AGS_FILE) : ags, 2058 kern_out); 2059 } 2060 } 2061 return (0); 2062 #undef RT 2063 } 2064 2065 2066 /* Update the kernel table to match the daemon table. */ 2067 static void 2068 fix_kern(void) 2069 { 2070 int i; 2071 struct khash *k, *pk, *knext; 2072 2073 2074 need_kern = age_timer; 2075 2076 /* Walk daemon table, updating the copy of the kernel table. */ 2077 (void) rn_walktree(rhead, walk_kern, NULL); 2078 ag_flush(0, 0, kern_out); 2079 2080 for (i = 0; i < KHASH_SIZE; i++) { 2081 pk = NULL; 2082 for (k = khash_bins[i]; k != NULL; k = knext) { 2083 knext = k->k_next; 2084 2085 /* Do not touch local interface routes */ 2086 if ((k->k_state & KS_DEPRE_IF) || 2087 (k->k_state & (KS_IF|KS_PASSIVE)) == KS_IF) { 2088 pk = k; 2089 continue; 2090 } 2091 2092 /* Do not touch static routes */ 2093 if (k->k_state & KS_STATIC) { 2094 kern_check_static(k, 0); 2095 pk = k; 2096 continue; 2097 } 2098 2099 /* check hold on routes deleted by the operator */ 2100 if (k->k_keep > now.tv_sec) { 2101 /* ensure we check when the hold is over */ 2102 LIM_SEC(need_kern, k->k_keep); 2103 pk = k; 2104 continue; 2105 } 2106 2107 if ((k->k_state & KS_DELETE) && 2108 !(k->k_state & KS_DYNAMIC)) { 2109 if ((k->k_dst == RIP_DEFAULT) && 2110 (k->k_ifp != NULL) && 2111 (kern_alternate(RIP_DEFAULT, 2112 k->k_mask, k->k_gate, k->k_ifp, 2113 NULL) == NULL)) 2114 rdisc_restore(k->k_ifp); 2115 kern_ioctl(k, RTM_DELETE, 0); 2116 if (pk != NULL) 2117 pk->k_next = knext; 2118 else 2119 khash_bins[i] = knext; 2120 free(k); 2121 continue; 2122 } 2123 2124 if (k->k_state & KS_DEL_ADD) 2125 kern_ioctl(k, RTM_DELETE, 0); 2126 2127 if (k->k_state & KS_ADD) { 2128 if ((k->k_dst == RIP_DEFAULT) && 2129 (k->k_ifp != NULL)) 2130 rdisc_suppress(k->k_ifp); 2131 kern_ioctl(k, RTM_ADD, 2132 ((0 != (k->k_state & (KS_GATEWAY | 2133 KS_DYNAMIC))) ? RTF_GATEWAY : 0)); 2134 } else if (k->k_state & KS_CHANGE) { 2135 /* 2136 * Should be using RTM_CHANGE here, but 2137 * since RTM_CHANGE is currently 2138 * not multipath-aware, and assumes 2139 * that RTF_GATEWAY implies the gateway 2140 * of the route for dst has to be 2141 * changed, we play safe, and do a del + add. 2142 */ 2143 kern_ioctl(k, RTM_DELETE, 0); 2144 kern_ioctl(k, RTM_ADD, 2145 ((0 != (k->k_state & (KS_GATEWAY | 2146 KS_DYNAMIC))) ? RTF_GATEWAY : 0)); 2147 } 2148 k->k_state &= ~(KS_ADD|KS_CHANGE|KS_DEL_ADD); 2149 2150 /* 2151 * Mark this route to be deleted in the next cycle. 2152 * This deletes routes that disappear from the 2153 * daemon table, since the normal aging code 2154 * will clear the bit for routes that have not 2155 * disappeared from the daemon table. 2156 */ 2157 k->k_state |= KS_DELETE; 2158 pk = k; 2159 } 2160 } 2161 } 2162 2163 2164 /* Delete a static route in the image of the kernel table. */ 2165 void 2166 del_static(in_addr_t dst, in_addr_t mask, in_addr_t gate, 2167 struct interface *ifp, int gone) 2168 { 2169 struct khash *k; 2170 struct rt_entry *rt; 2171 2172 /* 2173 * Just mark it in the table to be deleted next time the kernel 2174 * table is updated. 2175 * If it has already been deleted, mark it as such, and set its 2176 * keep-timer so that it will not be deleted again for a while. 2177 * This lets the operator delete a route added by the daemon 2178 * and add a replacement. 2179 */ 2180 k = kern_find(dst, mask, gate, ifp, NULL); 2181 if (k != NULL && (gate == 0 || k->k_gate == gate)) { 2182 k->k_state &= ~(KS_STATIC | KS_DYNAMIC | KS_CHECK); 2183 k->k_state |= KS_DELETE; 2184 if (gone) { 2185 k->k_state |= KS_DELETED; 2186 k->k_keep = now.tv_sec + K_KEEP_LIM; 2187 } 2188 } 2189 2190 rt = rtget(dst, mask); 2191 if (rt != NULL && (rt->rt_state & RS_STATIC)) 2192 rtbad(rt, NULL); 2193 } 2194 2195 2196 /* 2197 * Delete all routes generated from ICMP Redirects that use a given gateway, 2198 * as well as old redirected routes. 2199 */ 2200 void 2201 del_redirects(in_addr_t bad_gate, time_t old) 2202 { 2203 int i; 2204 struct khash *k; 2205 boolean_t dosupply = should_supply(NULL); 2206 2207 for (i = 0; i < KHASH_SIZE; i++) { 2208 for (k = khash_bins[i]; k != NULL; k = k->k_next) { 2209 if (!(k->k_state & KS_DYNAMIC) || 2210 (k->k_state & (KS_STATIC|KS_IF|KS_DEPRE_IF))) 2211 continue; 2212 2213 if (k->k_gate != bad_gate && k->k_redirect_time > old && 2214 !dosupply) 2215 continue; 2216 2217 k->k_state |= KS_DELETE; 2218 k->k_state &= ~KS_DYNAMIC; 2219 need_kern.tv_sec = now.tv_sec; 2220 trace_act("mark redirected %s --> %s for deletion", 2221 addrname(k->k_dst, k->k_mask, 0), 2222 naddr_ntoa(k->k_gate)); 2223 } 2224 } 2225 } 2226 2227 /* Start the daemon tables. */ 2228 void 2229 rtinit(void) 2230 { 2231 int i; 2232 struct ag_info *ag; 2233 2234 /* Initialize the radix trees */ 2235 rn_init(); 2236 (void) rn_inithead((void**)&rhead, 32); 2237 2238 /* mark all of the slots in the table free */ 2239 ag_avail = ag_slots; 2240 for (ag = ag_slots, i = 1; i < NUM_AG_SLOTS; i++) { 2241 ag->ag_fine = ag+1; 2242 ag++; 2243 } 2244 } 2245 2246 2247 static struct sockaddr_in dst_sock = {AF_INET}; 2248 static struct sockaddr_in mask_sock = {AF_INET}; 2249 2250 2251 static void 2252 set_need_flash(void) 2253 { 2254 if (!need_flash) { 2255 need_flash = _B_TRUE; 2256 /* 2257 * Do not send the flash update immediately. Wait a little 2258 * while to hear from other routers. 2259 */ 2260 no_flash.tv_sec = now.tv_sec + MIN_WAITTIME; 2261 } 2262 } 2263 2264 2265 /* Get a particular routing table entry */ 2266 struct rt_entry * 2267 rtget(in_addr_t dst, in_addr_t mask) 2268 { 2269 struct rt_entry *rt; 2270 2271 dst_sock.sin_addr.s_addr = dst; 2272 mask_sock.sin_addr.s_addr = htonl(mask); 2273 rt = (struct rt_entry *)rhead->rnh_lookup(&dst_sock, &mask_sock, rhead); 2274 if (rt == NULL || rt->rt_dst != dst || rt->rt_mask != mask) 2275 return (NULL); 2276 2277 return (rt); 2278 } 2279 2280 2281 /* Find a route to dst as the kernel would. */ 2282 struct rt_entry * 2283 rtfind(in_addr_t dst) 2284 { 2285 dst_sock.sin_addr.s_addr = dst; 2286 return ((struct rt_entry *)rhead->rnh_matchaddr(&dst_sock, rhead)); 2287 } 2288 2289 2290 /* add a route to the table */ 2291 void 2292 rtadd(in_addr_t dst, 2293 in_addr_t mask, 2294 uint16_t state, /* rt_state for the entry */ 2295 struct rt_spare *new) 2296 { 2297 struct rt_entry *rt; 2298 in_addr_t smask; 2299 int i; 2300 struct rt_spare *rts; 2301 2302 /* This is the only function that increments total_routes. */ 2303 if (total_routes == MAX_ROUTES) { 2304 msglog("have maximum (%d) routes", total_routes); 2305 return; 2306 } 2307 2308 rt = rtmalloc(sizeof (*rt), "rtadd"); 2309 (void) memset(rt, 0, sizeof (*rt)); 2310 rt->rt_spares = rtmalloc(SPARE_INC * sizeof (struct rt_spare), 2311 "rtadd"); 2312 rt->rt_num_spares = SPARE_INC; 2313 (void) memset(rt->rt_spares, 0, SPARE_INC * sizeof (struct rt_spare)); 2314 for (rts = rt->rt_spares, i = rt->rt_num_spares; i != 0; i--, rts++) 2315 rts->rts_metric = HOPCNT_INFINITY; 2316 2317 rt->rt_nodes->rn_key = (uint8_t *)&rt->rt_dst_sock; 2318 rt->rt_dst = dst; 2319 rt->rt_dst_sock.sin_family = AF_INET; 2320 if (mask != HOST_MASK) { 2321 smask = std_mask(dst); 2322 if ((smask & ~mask) == 0 && mask > smask) 2323 state |= RS_SUBNET; 2324 } 2325 mask_sock.sin_addr.s_addr = htonl(mask); 2326 rt->rt_mask = mask; 2327 rt->rt_spares[0] = *new; 2328 rt->rt_state = state; 2329 rt->rt_time = now.tv_sec; 2330 rt->rt_poison_metric = HOPCNT_INFINITY; 2331 rt->rt_seqno = update_seqno; 2332 2333 if (TRACEACTIONS) 2334 trace_add_del("Add", rt); 2335 2336 need_kern.tv_sec = now.tv_sec; 2337 set_need_flash(); 2338 2339 if (NULL == rhead->rnh_addaddr(&rt->rt_dst_sock, &mask_sock, rhead, 2340 rt->rt_nodes)) { 2341 msglog("rnh_addaddr() failed for %s mask=%s", 2342 naddr_ntoa(dst), naddr_ntoa(htonl(mask))); 2343 free(rt); 2344 } 2345 2346 total_routes++; 2347 } 2348 2349 2350 /* notice a changed route */ 2351 void 2352 rtchange(struct rt_entry *rt, 2353 uint16_t state, /* new state bits */ 2354 struct rt_spare *new, 2355 char *label) 2356 { 2357 if (rt->rt_metric != new->rts_metric) { 2358 /* 2359 * Fix the kernel immediately if it seems the route 2360 * has gone bad, since there may be a working route that 2361 * aggregates this route. 2362 */ 2363 if (new->rts_metric == HOPCNT_INFINITY) { 2364 need_kern.tv_sec = now.tv_sec; 2365 if (new->rts_time >= now.tv_sec - EXPIRE_TIME) 2366 new->rts_time = now.tv_sec - EXPIRE_TIME; 2367 } 2368 rt->rt_seqno = update_seqno; 2369 set_need_flash(); 2370 } 2371 2372 if (rt->rt_gate != new->rts_gate) { 2373 need_kern.tv_sec = now.tv_sec; 2374 rt->rt_seqno = update_seqno; 2375 set_need_flash(); 2376 } 2377 2378 state |= (rt->rt_state & RS_SUBNET); 2379 2380 /* Keep various things from deciding ageless routes are stale. */ 2381 if (!AGE_RT(state, rt->rt_spares[0].rts_origin, new->rts_ifp)) 2382 new->rts_time = now.tv_sec; 2383 2384 if (TRACEACTIONS) 2385 trace_change(rt, state, new, 2386 label ? label : "Chg "); 2387 2388 rt->rt_state = state; 2389 /* 2390 * If the interface state of the new primary route is good, 2391 * turn off RS_BADIF flag 2392 */ 2393 if ((rt->rt_state & RS_BADIF) && 2394 IS_IFF_UP(new->rts_ifp->int_if_flags) && 2395 !(new->rts_ifp->int_state & (IS_BROKE | IS_SICK))) 2396 rt->rt_state &= ~(RS_BADIF); 2397 2398 rt->rt_spares[0] = *new; 2399 } 2400 2401 2402 /* check for a better route among the spares */ 2403 static struct rt_spare * 2404 rts_better(struct rt_entry *rt) 2405 { 2406 struct rt_spare *rts, *rts1; 2407 int i; 2408 2409 /* find the best alternative among the spares */ 2410 rts = rt->rt_spares+1; 2411 for (i = rt->rt_num_spares, rts1 = rts+1; i > 2; i--, rts1++) { 2412 if (BETTER_LINK(rt, rts1, rts)) 2413 rts = rts1; 2414 } 2415 2416 return (rts); 2417 } 2418 2419 2420 /* switch to a backup route */ 2421 void 2422 rtswitch(struct rt_entry *rt, 2423 struct rt_spare *rts) 2424 { 2425 struct rt_spare swap; 2426 char label[10]; 2427 2428 /* Do not change permanent routes */ 2429 if (0 != (rt->rt_state & (RS_MHOME | RS_STATIC | 2430 RS_NET_SYN | RS_IF))) 2431 return; 2432 2433 /* find the best alternative among the spares */ 2434 if (rts == NULL) 2435 rts = rts_better(rt); 2436 2437 /* Do not bother if it is not worthwhile. */ 2438 if (!BETTER_LINK(rt, rts, rt->rt_spares)) 2439 return; 2440 2441 swap = rt->rt_spares[0]; 2442 (void) snprintf(label, sizeof (label), "Use #%d", 2443 (int)(rts - rt->rt_spares)); 2444 rtchange(rt, rt->rt_state & ~(RS_NET_SYN), rts, label); 2445 2446 if (swap.rts_metric == HOPCNT_INFINITY) { 2447 *rts = rts_empty; 2448 } else { 2449 *rts = swap; 2450 } 2451 2452 } 2453 2454 2455 void 2456 rtdelete(struct rt_entry *rt) 2457 { 2458 struct rt_entry *deleted_rt; 2459 struct rt_spare *rts; 2460 int i; 2461 in_addr_t gate = rt->rt_gate; /* for debugging */ 2462 2463 if (TRACEACTIONS) 2464 trace_add_del("Del", rt); 2465 2466 for (i = 0; i < rt->rt_num_spares; i++) { 2467 rts = &rt->rt_spares[i]; 2468 rts_delete(rt, rts); 2469 } 2470 2471 dst_sock.sin_addr.s_addr = rt->rt_dst; 2472 mask_sock.sin_addr.s_addr = htonl(rt->rt_mask); 2473 if (rt != (deleted_rt = 2474 ((struct rt_entry *)rhead->rnh_deladdr(&dst_sock, &mask_sock, 2475 rhead)))) { 2476 msglog("rnh_deladdr(%s) failed; found rt 0x%lx", 2477 rtname(rt->rt_dst, rt->rt_mask, gate), deleted_rt); 2478 if (deleted_rt != NULL) 2479 free(deleted_rt); 2480 } 2481 total_routes--; 2482 free(rt->rt_spares); 2483 free(rt); 2484 2485 if (dst_sock.sin_addr.s_addr == RIP_DEFAULT) { 2486 /* 2487 * we just deleted the default route. Trigger rdisc_sort 2488 * so that we can recover from any rdisc information that 2489 * is valid 2490 */ 2491 rdisc_timer.tv_sec = 0; 2492 } 2493 } 2494 2495 void 2496 rts_delete(struct rt_entry *rt, struct rt_spare *rts) 2497 { 2498 struct khash *k; 2499 2500 trace_upslot(rt, rts, &rts_empty); 2501 k = kern_find(rt->rt_dst, rt->rt_mask, 2502 rts->rts_gate, rts->rts_ifp, NULL); 2503 if (k != NULL && 2504 !(k->k_state & KS_DEPRE_IF) && 2505 ((k->k_state & (KS_IF|KS_PASSIVE)) != KS_IF)) { 2506 k->k_state |= KS_DELETE; 2507 need_kern.tv_sec = now.tv_sec; 2508 } 2509 2510 *rts = rts_empty; 2511 } 2512 2513 /* 2514 * Get rid of a bad route, and try to switch to a replacement. 2515 * If the route has gone bad because of a bad interface, 2516 * the information about the dead interface is available in badifp 2517 * for the purpose of sanity checks, if_flags checks etc. 2518 */ 2519 static void 2520 rtbad(struct rt_entry *rt, struct interface *badifp) 2521 { 2522 struct rt_spare new; 2523 uint16_t rt_state; 2524 2525 2526 if (badifp == NULL || (rt->rt_spares[0].rts_ifp == badifp)) { 2527 /* Poison the route */ 2528 new = rt->rt_spares[0]; 2529 new.rts_metric = HOPCNT_INFINITY; 2530 rt_state = rt->rt_state & ~(RS_IF | RS_LOCAL | RS_STATIC); 2531 } 2532 2533 if (badifp != NULL) { 2534 /* 2535 * Dont mark the rtentry bad unless the ifp for the primary 2536 * route is the bad ifp 2537 */ 2538 if (rt->rt_spares[0].rts_ifp != badifp) 2539 return; 2540 /* 2541 * badifp has just gone bad. We want to keep this 2542 * rt_entry around so that we tell our rip-neighbors 2543 * about the bad route, but we can't do anything 2544 * to the kernel itself, so mark it as RS_BADIF 2545 */ 2546 trace_misc("rtbad:Setting RS_BADIF (%s)", badifp->int_name); 2547 rt_state |= RS_BADIF; 2548 new.rts_ifp = &dummy_ifp; 2549 } 2550 rtchange(rt, rt_state, &new, 0); 2551 rtswitch(rt, 0); 2552 } 2553 2554 2555 /* 2556 * Junk a RS_NET_SYN or RS_LOCAL route, 2557 * unless it is needed by another interface. 2558 */ 2559 void 2560 rtbad_sub(struct rt_entry *rt, struct interface *badifp) 2561 { 2562 struct interface *ifp, *ifp1; 2563 struct intnet *intnetp; 2564 uint_t state; 2565 2566 2567 ifp1 = NULL; 2568 state = 0; 2569 2570 if (rt->rt_state & RS_LOCAL) { 2571 /* 2572 * Is this the route through loopback for the interface? 2573 * If so, see if it is used by any other interfaces, such 2574 * as a point-to-point interface with the same local address. 2575 */ 2576 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2577 /* Retain it if another interface needs it. */ 2578 if (ifp->int_addr == rt->rt_ifp->int_addr) { 2579 state |= RS_LOCAL; 2580 ifp1 = ifp; 2581 break; 2582 } 2583 } 2584 2585 } 2586 2587 if (!(state & RS_LOCAL)) { 2588 /* 2589 * Retain RIPv1 logical network route if there is another 2590 * interface that justifies it. 2591 */ 2592 if (rt->rt_state & RS_NET_SYN) { 2593 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2594 if ((ifp->int_state & IS_NEED_NET_SYN) && 2595 rt->rt_mask == ifp->int_std_mask && 2596 rt->rt_dst == ifp->int_std_addr) { 2597 state |= RS_NET_SYN; 2598 ifp1 = ifp; 2599 break; 2600 } 2601 } 2602 } 2603 2604 /* or if there is an authority route that needs it. */ 2605 for (intnetp = intnets; intnetp != NULL; 2606 intnetp = intnetp->intnet_next) { 2607 if (intnetp->intnet_addr == rt->rt_dst && 2608 intnetp->intnet_mask == rt->rt_mask) { 2609 state |= (RS_NET_SYN | RS_NET_INT); 2610 break; 2611 } 2612 } 2613 } 2614 2615 if (ifp1 != NULL || (state & RS_NET_SYN)) { 2616 struct rt_spare new = rt->rt_spares[0]; 2617 new.rts_ifp = ifp1; 2618 rtchange(rt, ((rt->rt_state & ~(RS_NET_SYN|RS_LOCAL)) | state), 2619 &new, 0); 2620 } else { 2621 rtbad(rt, badifp); 2622 } 2623 } 2624 2625 /* 2626 * Called while walking the table looking for sick interfaces 2627 * or after a time change. 2628 */ 2629 int 2630 walk_bad(struct radix_node *rn, 2631 void *argp) 2632 { 2633 #define RT ((struct rt_entry *)rn) 2634 struct rt_spare *rts; 2635 int i, j = -1; 2636 2637 /* fix any spare routes through the interface */ 2638 for (i = 1; i < RT->rt_num_spares; i++) { 2639 rts = &((struct rt_entry *)rn)->rt_spares[i]; 2640 2641 if (rts->rts_metric < HOPCNT_INFINITY && 2642 (rts->rts_ifp == NULL || 2643 (rts->rts_ifp->int_state & IS_BROKE))) 2644 rts_delete(RT, rts); 2645 else { 2646 if (rts->rts_origin != RO_NONE) 2647 j = i; 2648 } 2649 } 2650 2651 /* 2652 * Deal with the main route 2653 * finished if it has been handled before or if its interface is ok 2654 */ 2655 if (RT->rt_ifp == NULL || !(RT->rt_ifp->int_state & IS_BROKE)) 2656 return (0); 2657 2658 /* Bad routes for other than interfaces are easy. */ 2659 if (!(RT->rt_state & (RS_IF | RS_NET_SYN | RS_LOCAL))) { 2660 if (j > 0) { 2661 RT->rt_spares[0].rts_metric = HOPCNT_INFINITY; 2662 rtswitch(RT, NULL); 2663 } else { 2664 rtbad(RT, (struct interface *)argp); 2665 } 2666 return (0); 2667 } 2668 2669 rtbad_sub(RT, (struct interface *)argp); 2670 return (0); 2671 #undef RT 2672 } 2673 2674 /* 2675 * Called while walking the table to replace a duplicate interface 2676 * with a backup. 2677 */ 2678 int 2679 walk_rewire(struct radix_node *rn, void *argp) 2680 { 2681 struct rt_entry *RT = (struct rt_entry *)rn; 2682 struct rewire_data *wire = (struct rewire_data *)argp; 2683 struct rt_spare *rts; 2684 int i; 2685 2686 /* fix any spare routes through the interface */ 2687 rts = RT->rt_spares; 2688 for (i = RT->rt_num_spares; i > 0; i--, rts++) { 2689 if (rts->rts_ifp == wire->if_old) { 2690 rts->rts_ifp = wire->if_new; 2691 if ((RT->rt_dst == RIP_DEFAULT) && 2692 (wire->if_old->int_state & IS_SUPPRESS_RDISC)) 2693 rdisc_suppress(rts->rts_ifp); 2694 if ((rts->rts_metric += wire->metric_delta) > 2695 HOPCNT_INFINITY) 2696 rts->rts_metric = HOPCNT_INFINITY; 2697 2698 /* 2699 * If the main route is getting a worse metric, 2700 * then it may be time to switch to a backup. 2701 */ 2702 if (i == RT->rt_num_spares && wire->metric_delta > 0) { 2703 rtswitch(RT, NULL); 2704 } 2705 } 2706 } 2707 2708 return (0); 2709 } 2710 2711 /* Check the age of an individual route. */ 2712 static int 2713 walk_age(struct radix_node *rn, void *argp) 2714 { 2715 #define RT ((struct rt_entry *)rn) 2716 struct interface *ifp; 2717 struct rt_spare *rts; 2718 int i; 2719 in_addr_t age_bad_gate = *(in_addr_t *)argp; 2720 2721 2722 /* 2723 * age all of the spare routes, including the primary route 2724 * currently in use 2725 */ 2726 rts = RT->rt_spares; 2727 for (i = RT->rt_num_spares; i != 0; i--, rts++) { 2728 2729 ifp = rts->rts_ifp; 2730 if (i == RT->rt_num_spares) { 2731 if (!AGE_RT(RT->rt_state, rts->rts_origin, ifp)) { 2732 /* 2733 * Keep various things from deciding ageless 2734 * routes are stale 2735 */ 2736 rts->rts_time = now.tv_sec; 2737 continue; 2738 } 2739 2740 /* forget RIP routes after RIP has been turned off. */ 2741 if (rip_sock < 0) { 2742 rts->rts_time = now_stale + 1; 2743 } 2744 } 2745 2746 /* age failing routes */ 2747 if (age_bad_gate == rts->rts_gate && 2748 rts->rts_time >= now_stale) { 2749 rts->rts_time -= SUPPLY_INTERVAL; 2750 } 2751 2752 /* trash the spare routes when they go bad */ 2753 if (rts->rts_origin == RO_RIP && 2754 ((rip_sock < 0) || 2755 (rts->rts_metric < HOPCNT_INFINITY && 2756 now_garbage > rts->rts_time)) && 2757 i != RT->rt_num_spares) { 2758 rts_delete(RT, rts); 2759 } 2760 } 2761 2762 2763 /* finished if the active route is still fresh */ 2764 if (now_stale <= RT->rt_time) 2765 return (0); 2766 2767 /* try to switch to an alternative */ 2768 rtswitch(RT, NULL); 2769 2770 /* Delete a dead route after it has been publically mourned. */ 2771 if (now_garbage > RT->rt_time) { 2772 rtdelete(RT); 2773 return (0); 2774 } 2775 2776 /* Start poisoning a bad route before deleting it. */ 2777 if (now.tv_sec - RT->rt_time > EXPIRE_TIME) { 2778 struct rt_spare new = RT->rt_spares[0]; 2779 2780 new.rts_metric = HOPCNT_INFINITY; 2781 rtchange(RT, RT->rt_state, &new, 0); 2782 } 2783 return (0); 2784 } 2785 2786 2787 /* Watch for dead routes and interfaces. */ 2788 void 2789 age(in_addr_t bad_gate) 2790 { 2791 struct interface *ifp; 2792 int need_query = 0; 2793 2794 /* 2795 * If not listening to RIP, there is no need to age the routes in 2796 * the table. 2797 */ 2798 age_timer.tv_sec = (now.tv_sec 2799 + ((rip_sock < 0) ? NEVER : SUPPLY_INTERVAL)); 2800 2801 /* 2802 * Check for dead IS_REMOTE interfaces by timing their 2803 * transmissions. 2804 */ 2805 for (ifp = ifnet; ifp; ifp = ifp->int_next) { 2806 if (!(ifp->int_state & IS_REMOTE)) 2807 continue; 2808 2809 /* ignore unreachable remote interfaces */ 2810 if (!check_remote(ifp)) 2811 continue; 2812 2813 /* Restore remote interface that has become reachable */ 2814 if (ifp->int_state & IS_BROKE) 2815 if_ok(ifp, "remote ", _B_FALSE); 2816 2817 if (ifp->int_act_time != NEVER && 2818 now.tv_sec - ifp->int_act_time > EXPIRE_TIME) { 2819 writelog(LOG_NOTICE, 2820 "remote interface %s to %s timed out after" 2821 " %ld:%ld", 2822 ifp->int_name, 2823 naddr_ntoa(ifp->int_dstaddr), 2824 (now.tv_sec - ifp->int_act_time)/60, 2825 (now.tv_sec - ifp->int_act_time)%60); 2826 if_sick(ifp, _B_FALSE); 2827 } 2828 2829 /* 2830 * If we have not heard from the other router 2831 * recently, ask it. 2832 */ 2833 if (now.tv_sec >= ifp->int_query_time) { 2834 ifp->int_query_time = NEVER; 2835 need_query = 1; 2836 } 2837 } 2838 2839 /* Age routes. */ 2840 (void) rn_walktree(rhead, walk_age, &bad_gate); 2841 2842 /* 2843 * delete old redirected routes to keep the kernel table small 2844 * and prevent blackholes 2845 */ 2846 del_redirects(bad_gate, now.tv_sec-STALE_TIME); 2847 2848 /* Update the kernel routing table. */ 2849 fix_kern(); 2850 2851 /* poke reticent remote gateways */ 2852 if (need_query) 2853 rip_query(); 2854 } 2855 2856 void 2857 kern_dump(void) 2858 { 2859 int i; 2860 struct khash *k; 2861 2862 for (i = 0; i < KHASH_SIZE; i++) { 2863 for (k = khash_bins[i]; k != NULL; k = k->k_next) 2864 trace_khash(k); 2865 } 2866 } 2867 2868 2869 static struct interface * 2870 gwkludge_iflookup(in_addr_t dstaddr, in_addr_t addr, in_addr_t mask) 2871 { 2872 uint32_t int_state; 2873 struct interface *ifp; 2874 2875 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2876 int_state = ifp->int_state; 2877 2878 if (!(int_state & IS_REMOTE)) 2879 continue; 2880 2881 if (ifp->int_dstaddr == dstaddr && ifp->int_addr == addr && 2882 ifp->int_mask == mask) 2883 return (ifp); 2884 } 2885 return (NULL); 2886 } 2887